gchhablani commited on
Commit
66503d3
1 Parent(s): b25e52a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -4
README.md CHANGED
@@ -17,7 +17,7 @@ Disclaimer: This model card has been written by [gchhablani](https://huggingface
17
 
18
  ## Model description
19
 
20
- FNet is a transformers model with attention replaced with fourier transforms. It is pretrained on a large corpus of
21
  English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling
22
  them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and
23
  labels from those texts. More precisely, it was pretrained with two objectives:
@@ -46,9 +46,10 @@ to make decisions, such as sequence classification, token classification or ques
46
  generation you should look at model like GPT2.
47
 
48
  ### How to use
49
-
50
  You can use this model directly with a pipeline for masked language modeling:
51
 
 
 
52
  ```python
53
  >>> from transformers import FNetForMaskedLM, FNetTokenizer, pipeline
54
  >>> tokenizer = FNetTokenizer.from_pretrained("google/fnet-base")
@@ -68,12 +69,14 @@ You can use this model directly with a pipeline for masked language modeling:
68
 
69
  Here is how to use this model to get the features of a given text in PyTorch:
70
 
 
 
71
  ```python
72
  from transformers import FNetTokenizer, FNetModel
73
  tokenizer = FNetTokenizer.from_pretrained("google/fnet-base")
74
  model = FNetModel.from_pretrained("google/fnet-base")
75
  text = "Replace me by any text you'd like."
76
- encoded_input = tokenizer(text, return_tensors='pt')
77
  output = model(**encoded_input)
78
  ```
79
 
@@ -172,4 +175,7 @@ Glue test results:
172
  biburl = {https://dblp.org/rec/journals/corr/abs-2105-03824.bib},
173
  bibsource = {dblp computer science bibliography, https://dblp.org}
174
  }
175
- ```
 
 
 
 
17
 
18
  ## Model description
19
 
20
+ FNet is a transformers model with attention replaced with fourier transforms. Hence, the inputs do not contain an `attention_mask`. It is pretrained on a large corpus of
21
  English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling
22
  them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and
23
  labels from those texts. More precisely, it was pretrained with two objectives:
 
46
  generation you should look at model like GPT2.
47
 
48
  ### How to use
 
49
  You can use this model directly with a pipeline for masked language modeling:
50
 
51
+ **Note: The mask filling pipeline doesn't work exactly as the original model performs masking after converting to tokens. In masking pipeline an additional space is added after the [MASK].**
52
+
53
  ```python
54
  >>> from transformers import FNetForMaskedLM, FNetTokenizer, pipeline
55
  >>> tokenizer = FNetTokenizer.from_pretrained("google/fnet-base")
 
69
 
70
  Here is how to use this model to get the features of a given text in PyTorch:
71
 
72
+ **Note: You must specify the maximum sequence length to be 512 and truncate/pad to the same length because the original model has no attention mask and considers all the hidden states during forward pass.**
73
+
74
  ```python
75
  from transformers import FNetTokenizer, FNetModel
76
  tokenizer = FNetTokenizer.from_pretrained("google/fnet-base")
77
  model = FNetModel.from_pretrained("google/fnet-base")
78
  text = "Replace me by any text you'd like."
79
+ encoded_input = tokenizer(text, return_tensors='pt', padding='max_length', truncation=True, max_length=512)
80
  output = model(**encoded_input)
81
  ```
82
 
 
175
  biburl = {https://dblp.org/rec/journals/corr/abs-2105-03824.bib},
176
  bibsource = {dblp computer science bibliography, https://dblp.org}
177
  }
178
+ ```
179
+
180
+ ## Contributions
181
+ Thanks to @gchhablani for adding this model.