hdallatorre
commited on
Commit
•
73917d6
1
Parent(s):
ff80424
Update README.md
Browse files
README.md
CHANGED
@@ -120,7 +120,7 @@ The model was trained on a DGXH100 node with 8 GPUs on a total of 23B tokens for
|
|
120 |
|
121 |
### Architecture
|
122 |
|
123 |
-
The model is composed of the [nucleotide-transformer-v2-
|
124 |
the language model head and replaced it by a 1-dimensional U-Net segmentation head [4] made of 2 downsampling convolutional blocks and 2 upsampling convolutional blocks. Each of these
|
125 |
blocks is made of 2 convolutional layers with 1, 024 and 2, 048 kernels respectively. This additional segmentation head accounts for 53 million parameters, bringing the total number of parameters
|
126 |
to 562M.
|
|
|
120 |
|
121 |
### Architecture
|
122 |
|
123 |
+
The model is composed of the [nucleotide-transformer-v2-500m-multi-species](https://huggingface.co/InstaDeepAI/nucleotide-transformer-v2-500m-multi-species) encoder, from which we removed
|
124 |
the language model head and replaced it by a 1-dimensional U-Net segmentation head [4] made of 2 downsampling convolutional blocks and 2 upsampling convolutional blocks. Each of these
|
125 |
blocks is made of 2 convolutional layers with 1, 024 and 2, 048 kernels respectively. This additional segmentation head accounts for 53 million parameters, bringing the total number of parameters
|
126 |
to 562M.
|