yuchen005 commited on
Commit
66619d3
1 Parent(s): 7706f64

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +13 -7
README.md CHANGED
@@ -27,25 +27,31 @@ This repo releases the Robust HyPoradise dataset in paper "Large Language Models
27
 
28
  **UPDATE (Apr-18-2024):** We have released the training data, which follows the same format as test data.
29
  Considering the file size, the uploaded training data does not contain the speech features (vast size).
30
- Alternatively, we have provided a script named `add_speech_feats_to_train_data.py` to generate them from raw speech (.wav).
31
  You need to specify the raw speech path from utterance id in the script.
32
  Here are the available speech data: [CHiME-4](https://entuedu-my.sharepoint.com/:f:/g/personal/yuchen005_e_ntu_edu_sg/EuLgMQbjrIJHk7dKPkjcDMIB4SYgXKKP8VBxyiZk3qgdgA),
33
  [VB-DEMAND](https://datashare.ed.ac.uk/handle/10283/2791), [LS-FreeSound](https://github.com/archiki/Robust-E2E-ASR), [NOIZEUS](https://ecs.utdallas.edu/loizou/speech/noizeus/).
34
 
35
- **IMPORTANT:** The vast speech feature size mentioned above is because Whisper requires a fix input length of 30s that is too long. Please do the follwing step before running data generation:
36
- - Modified the [model code](https://github.com/openai/whisper/blob/main/whisper/model.py#L167) `x = (x + self.positional_embedding).to(x.dtype)` to be `x = (x + self.positional_embedding[:x.shape[1], :]).to(x.dtype)`
37
 
38
- **UPDATE (Apr-29-2024):** To support customization, We release the script `generate_robust_hp.py` for users to generate train/test data from their own ASR datasets.
39
- We also release two necessary packages for generation, one is the `jiwer` package that is locally imported in `generate_robust_hp.py`, another one is the whisper decoding script `decoding.py` that should be put under locally installed whisper directory `<your-path>/whisper/whisper`.
 
 
 
 
 
 
 
 
40
 
41
 
42
  If you consider this work would be related or useful for your research, please kindly consider to cite the work in ICLR 2024. Thank you.
43
 
44
- ```bib
45
  @inproceedings{hu2024large,
46
  title={Large Language Models are Efficient Learners of Noise-Robust Speech Recognition},
47
  author={Hu, Yuchen and Chen, Chen and Yang, Chao-Han Huck and Li, Ruizhe and Zhang, Chao and Chen, Pin-Yu and Chng, Eng Siong},
48
  booktitle={International Conference on Learning Representations},
49
  year={2024}
50
  }
51
- ```
 
27
 
28
  **UPDATE (Apr-18-2024):** We have released the training data, which follows the same format as test data.
29
  Considering the file size, the uploaded training data does not contain the speech features (vast size).
30
+ Alternatively, we have provided a script named ***add_speech_feats_to_train_data.py*** to generate them from raw speech (.wav).
31
  You need to specify the raw speech path from utterance id in the script.
32
  Here are the available speech data: [CHiME-4](https://entuedu-my.sharepoint.com/:f:/g/personal/yuchen005_e_ntu_edu_sg/EuLgMQbjrIJHk7dKPkjcDMIB4SYgXKKP8VBxyiZk3qgdgA),
33
  [VB-DEMAND](https://datashare.ed.ac.uk/handle/10283/2791), [LS-FreeSound](https://github.com/archiki/Robust-E2E-ASR), [NOIZEUS](https://ecs.utdallas.edu/loizou/speech/noizeus/).
34
 
 
 
35
 
36
+ **IMPORTANT:** The vast speech feature size mentioned above is because Whisper requires a fix input length of 30s that is too long. Please do the follwing step to remove it before running ***add_speech_feats_to_train_data.py***:
37
+ - Modified the [whisper model code](https://github.com/openai/whisper/blob/main/whisper/model.py#L167) "x = (x + self.positional_embedding).to(x.dtype)" to be "x = (x + self.positional_embedding[:x.shape[1], :]).to(x.dtype)"
38
+
39
+
40
+ **UPDATE (Apr-29-2024):** To support customization, We release the script ***generate_robust_hp.py*** for users to generate train/test data from their own ASR datasets.
41
+ We also release two necessary packages for generation: "my_jiwer" and "decoding.py".
42
+ To summary, you will need to do the following three steps before running ***generate_robust_hp.py***:
43
+ - Modified the [whisper model code](https://github.com/openai/whisper/blob/main/whisper/model.py#L167) "x = (x + self.positional_embedding).to(x.dtype)" to be "x = (x + self.positional_embedding[:x.shape[1], :]).to(x.dtype)"
44
+ - Specify the absolute path of "my_jiwer" directory in ***generate_robust_hp.py*** (sys.path.append)
45
+ - Put our whisper decoding script "decoding.py" under your locally installed whisper directory "<your-path>/whisper/whisper"
46
 
47
 
48
  If you consider this work would be related or useful for your research, please kindly consider to cite the work in ICLR 2024. Thank you.
49
 
50
+ """bib
51
  @inproceedings{hu2024large,
52
  title={Large Language Models are Efficient Learners of Noise-Robust Speech Recognition},
53
  author={Hu, Yuchen and Chen, Chen and Yang, Chao-Han Huck and Li, Ruizhe and Zhang, Chao and Chen, Pin-Yu and Chng, Eng Siong},
54
  booktitle={International Conference on Learning Representations},
55
  year={2024}
56
  }
57
+ """