repo_name
stringclasses 28
values | pr_number
int64 1.86k
122k
| pr_title
stringlengths 5
204
| author
stringlengths 3
58
| git_commit_prev
stringlengths 40
40
| git_commit_curr
stringlengths 40
40
| date_created
stringlengths 25
25
| date_merged
stringlengths 25
25
| query
stringlengths 12
65.6k
| context_file_path
stringlengths 6
233
| label
int64 -1
1
| language
stringclasses 5
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
keras-team/keras | 18,837 | Add keras.backend.device() API for device scope | qlzh727 | 06d31d9e06d74dcf932d830bbdf1e4bf1447bf87 | 1ebe1d052e27166ebcd4fe8ffaa6fb7637c97d8f | 2023-11-27 23:51:38+00:00 | 2023-11-29 19:33:56+00:00 | Add keras.backend.device() API for device scope. | ./keras/layers/preprocessing/index_lookup_test.py | -1 | python |
keras-team/keras | 18,837 | Add keras.backend.device() API for device scope | qlzh727 | 06d31d9e06d74dcf932d830bbdf1e4bf1447bf87 | 1ebe1d052e27166ebcd4fe8ffaa6fb7637c97d8f | 2023-11-27 23:51:38+00:00 | 2023-11-29 19:33:56+00:00 | Add keras.backend.device() API for device scope. | ./keras/saving/serialization_lib.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/ops/numpy_test.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/ops/numpy.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/tensorflow/numpy.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/numpy/numpy.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/jax/numpy.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/torch/numpy.py | 1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/activations/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/callbacks/model_checkpoint_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/pooling/max_pooling_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/torch/optimizers/torch_adadelta.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/optimizers/optimizer.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/saving/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/preprocessing/random_contrast_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./conftest.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/datasets/imdb.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/models/variable_mapping.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/reshaping/cropping2d_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/pooling/max_pooling3d.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/pooling/average_pooling1d.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./benchmarks/layer_benchmark/rnn_benchmark.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/random/random.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/activations/prelu_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/callbacks/callback_list.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/core/masking_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/optimizers/adamw_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/tensorflow/nn.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/convolutional/base_separable_conv.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/trainers/trainer_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/pooling/max_pooling1d.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./guides/understanding_masking_and_padding.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/models/functional_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/rnn/simple_rnn_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/vision/mlp_image_classification.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/vision/image_classification_with_vision_transformer.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/attention/grouped_query_attention_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/tensorflow/structured_data/movielens_recommendations_transformers.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/preprocessing/hashed_crossing.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/normalization/unit_normalization.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/trainers/data_adapters/tf_dataset_adapter_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/common/variables_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./benchmarks/layer_benchmark/attention_benchmark.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/normalization/layer_normalization_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/utils/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/legacy/saving/saving_utils.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/utils/tracking_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/jax/math.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/pytorch/torchvision_keras.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/reshaping/up_sampling1d.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/metrics/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/timeseries/timeseries_weather_forecasting.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/preprocessing/random_crop.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/tensorflow/keras_recipes/antirectifier.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./benchmarks/layer_benchmark/__init__.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/tensorflow/trainer.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/testing/test_case.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/export/export_lib.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/layers/preprocessing/integer_lookup_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./integration_tests/basic_full_flow.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/utils/nest.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/initializers/random_initializers_test.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/torch/math.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/tensorflow/generative/wgan_gp.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/backend/numpy/nn.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./keras/applications/inception_resnet_v2.py | -1 | python |
keras-team/keras | 18,831 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where` | james77777778 | 866b745ebdafb29248f6a0946fbed1ce1cbcba90 | 7a6502200efe7c8b9f9cc1f82b05b2fba0199dc1 | 2023-11-26 18:46:54+00:00 | 2023-11-27 16:32:55+00:00 | Apply `backend.result_type` to `cumprod`, `cumsum`, `nonzero`, `power`, `take`, `take_along_axis`, `tensordot`, `tile`, `trace`, `transpose`, `tril`, `triu`, `vdot`, `vstack`, `where`. This PR should be the final one for `ops.numpy` dtype inference. Now, all ops (excluding `ndim`, `size` and complex-related ones) have consistent dtype inference across all backends!
<details>
- [x] abs
- [x] absolute
- [x] add
- [x] all
- [x] amax
- [x] amin
- [x] append
- [x] arange
- [x] arccos
- [x] arccosh
- [x] arcsin
- [x] arcsinh
- [x] arctan
- [x] arctan2
- [x] arctanh
- [x] argmax
- [x] argmin
- [x] argsort
- [x] array
- [x] average
- [x] bincount
- [x] broadcast_to
- [x] ceil
- [x] clip
- [x] concatenate
- [ ] conj (Keras does not directly support complex data)
- [ ] conjugate (Keras does not directly support complex data)
- [x] copy
- [x] cos
- [x] cosh
- [x] count_nonzero
- [x] cross
- [x] cumprod
- [x] cumsum
- [x] diag
- [x] diagonal
- [x] diff
- [x] digitize
- [x] divide
- [x] dot
- [x] einsum
- [x] empty
- [x] equal
- [x] exp
- [x] expand_dims
- [x] expm1
- [x] eye
- [x] flip
- [x] floor
- [x] full
- [x] full_like
- [x] greater
- [x] greater_equal
- [x] hstack
- [x] identity
- [ ] imag (Keras does not directly support complex data)
- [ ] interp (Keras lacks this op)
- [x] isclose
- [x] isfinite
- [x] isinf
- [x] isnan
- [x] less
- [x] less_equal
- [x] linspace
- [x] log
- [x] log10
- [x] log1p
- [x] log2
- [x] logaddexp
- [x] logical_and
- [x] logical_not
- [x] logical_or
- [x] logspace
- [x] matmul
- [x] max
- [x] maximum
- [x] mean
- [x] median
- [x] meshgrid
- [ ] mgrid (Keras lacks this op)
- [x] min
- [x] minimum
- [x] mod
- [x] moveaxis
- [x] multiply
- [x] nan_to_num
- [ ] ndim
- [x] nonzero
- [x] not_equal
- [x] ones
- [x] ones_like
- [x] outer
- [x] pad
- [ ] percentile (Keras lacks this op)
- [x] power
- [x] prod
- [x] quantile
- [x] ravel
- [ ] real (Keras does not directly support complex data)
- [ ] reciprocal (Keras lacks this op)
- [x] repeat
- [x] reshape
- [x] roll
- [x] round
- [x] sign
- [x] sin
- [x] sinh
- [ ] size
- [x] sort
- [x] split
- [x] sqrt
- [x] square
- [x] squeeze
- [x] stack
- [x] std
- [x] subtract
- [x] sum
- [x] swapaxes
- [x] take
- [x] take_along_axis
- [x] tan
- [x] tanh
- [x] tensordot
- [x] tile
- [x] trace
- [x] transpose
- [x] tri
- [x] tril
- [x] triu
- [x] true_divide
- [x] vdot
- [x] vstack
- [x] where
- [x] zeros
- [x] zeros_like
</details>
| ./examples/keras_io/vision/siamese_contrastive.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/separable_conv2d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/conv2d_transpose.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/conv1d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/base_separable_conv.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/base_conv.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/depthwise_conv2d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/base_depthwise_conv.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/conv3d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/separable_conv1d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/depthwise_conv1d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/conv2d.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/convolutional/conv3d_transpose.py | 1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/normalization/spectral_normalization_test.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/trainers/data_adapters/data_adapter_utils.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/trainers/data_adapters/array_data_adapter.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./examples/keras_io/structured_data/tabtransformer.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/regularization/gaussian_dropout.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/utils/backend_utils.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/saving/serialization_lib.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/preprocessing/random_zoom_test.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./benchmarks/model_benchmark/__init__.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/merging/maximum.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/trainers/trainer.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/backend/torch/random.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/utils/io_utils.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/applications/densenet.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./benchmarks/layer_benchmark/activation_benchmark.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/optimizers/rmsprop.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./examples/keras_io/tensorflow/audio/speaker_recognition_using_cnn.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./guides/custom_train_step_in_torch.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/layers/normalization/spectral_normalization.py | -1 | python |
keras-team/keras | 18,826 | Change Conv layers Doc string wrt padding. | SuryanarayanaY | 64c21f2f9125c7ea75943d278ecd95ff18b9af7f | 03f4a4fe3c0b2be47681e98275c11cd4bc786c8f | 2023-11-24 01:28:31+00:00 | 2023-11-24 13:34:20+00:00 | Change Conv layers Doc string wrt padding.. "Updating the args documentation of padding='same' for all convolutional layers.
Reference issue #15703.
Reference PRs #15771 closed due to stale.
The documentation for padding in all conv layers listed above states that, "" ""same"" results in padding with zeros evenly to the left/right or up/down of the input such that output has the same height/width dimension as the input."". This seems incorrect as it is true only when `strides=1`. Hence update the same. | ./keras/backend/common/name_scope_test.py | -1 | python |