Add results and notices for results for GigaSpeech transducer & wavlm

matthewkperez · Oct 25, 2024 · d98e949 · d98e949
1 parent 231c78a
commit d98e949
Show file tree

Hide file tree

Showing 2 changed files with 14 additions and 2 deletions.
diff --git a/recipes/GigaSpeech/ASR/CTC/README.md b/recipes/GigaSpeech/ASR/CTC/README.md
@@ -62,7 +62,9 @@ This can be done by modifying the current recipe. We invite you to have a look a
 
 | Release | Hyperparams file | Decoding method | Finetuning Split | Test WER | Dev WER |  HuggingFace link | Full model link | Training GPUs |
 |:-------------:|:---------------------------:|  :----------:|  :-----:| :-----:| :-----:| :-----:| :-----:| :-----:|
-| 05-08-23 | train_hf_wavlm.yaml | GreedySearch | XL  | xx | xx | TBD | TBD | 4xRTX 3090 |
+| 25-10-2024 | train_hf_wavlm.yaml | GreedySearch | XL  | 11.88% | 11.86% | Unavailable\* | Unavailable\* | 8xRTX 3090 |
+
+\*: Unfortunately, we are unable to upload the checkpoints for the WavLM model at this time. We currently don't have plans to remedy this.
 
 # **Citing SpeechBrain**
 Please, cite SpeechBrain if you use it for your research or business.

diff --git a/recipes/GigaSpeech/ASR/transducer/README.md b/recipes/GigaSpeech/ASR/transducer/README.md
@@ -48,10 +48,18 @@ According to our tests, the performance is not affected.
 
 Results are obtained with beam search and no LM (no-streaming i.e. full context).
 
+**TBD: The final models are currently in training.** This model has already been succesfully trained, though. This will be updated when the checkpoints are ready for download.
+
+<!--
+
 | Language | Release |  LM | Val. CER | Val. WER | Test CER | Test WER | Model link | GPUs |
 | ------------- |:-------------:| -----:| -----:| -----:| -----:| -----:| :-----------:| :-----------:|
 
-The output folders with checkpoints and logs can be found [here](https://www.dropbox.com/sh/852eq7pbt6d65ai/AACv4wAzk1pWbDo4fjVKLICYa?dl=0).
+-->
+
+<!-- NOT READY YET: also update the following URL when uploaded
+
+The output folders with checkpoints and logs can be found [here](). -->
 
 ## Streaming model
 
@@ -74,6 +82,8 @@ may end up forming indirect dependencies to audio many seconds ago.
 |       | full | cs=32 (1280ms) | 16 (640ms) | 8 (320ms) |
 |:-----:|:----:|:-----:|:-----:|:-----:|
 
+**TBD: The final models are currently in training.** This model has already been succesfully trained, though. This will be updated when the checkpoints are ready for download.
+
 ### Inference
 
 Once your model is trained, you need a few manual steps in order to use it with the high-level streaming interfaces (`speechbrain.inference.ASR.StreamingASR`):