sanchit-gandhi HF staff commited on
Commit
98ac223
1 Parent(s): 72ddd48

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -2
README.md CHANGED
@@ -16,6 +16,9 @@ It achieves the following WER results on the evaluation set:
16
  - Normalised WER: 6.324
17
  - Orthographic WER: 8.233
18
 
 
 
 
19
  ## Model description
20
 
21
  We copy the entire encoder module and freeze it during training. We copy only two decoder layers, which are initialised from the first and last decoder layers from Whisper. All other decoder layers from Whisper are discarded.
@@ -28,7 +31,7 @@ The model was trained and evaluated on the German subset of the [Common Voice 15
28
 
29
  ## Training procedure
30
 
31
- To reproduce this training run, first clone and install Distil-Whisper according to the instructions [here].
32
 
33
  Next, we can pick a name for our distilled model, e.g. `distil-whisper-large-v3-de-kd`. We can then run the following command to create a repository under this name:
34
 
@@ -45,7 +48,7 @@ git lfs install
45
  git clone https://huggingface.co/sanchit-gandhi/distil-whisper-large-v3-de-kd
46
  ```
47
 
48
- > **Note:** Be sure to change the repo address to `https://huggingface.co/<your-user-name>/<your-repo-name>`
49
 
50
  Next, copy the relevant training scrips from Distil-Whisper to the repository:
51
 
 
16
  - Normalised WER: 6.324
17
  - Orthographic WER: 8.233
18
 
19
+ Full tensorboard logs can be found under the tab [Training Metrics](https://huggingface.co/sanchit-gandhi/distil-whisper-large-v3-de-kd/tensorboard?params=scalars#frame),
20
+ and steps to reproduce [here](https://huggingface.co/sanchit-gandhi/distil-whisper-large-v3-de-kd#training-procedure).
21
+
22
  ## Model description
23
 
24
  We copy the entire encoder module and freeze it during training. We copy only two decoder layers, which are initialised from the first and last decoder layers from Whisper. All other decoder layers from Whisper are discarded.
 
31
 
32
  ## Training procedure
33
 
34
+ To reproduce this training run, first clone and install Distil-Whisper according to the instructions [here](https://github.com/huggingface/distil-whisper/tree/main/training#requirements).
35
 
36
  Next, we can pick a name for our distilled model, e.g. `distil-whisper-large-v3-de-kd`. We can then run the following command to create a repository under this name:
37
 
 
48
  git clone https://huggingface.co/sanchit-gandhi/distil-whisper-large-v3-de-kd
49
  ```
50
 
51
+ **Note:** Be sure to change the repo address to `https://huggingface.co/<your-user-name>/<your-repo-name>`
52
 
53
  Next, copy the relevant training scrips from Distil-Whisper to the repository:
54