gchhablani commited on
Commit
3e3e3eb
1 Parent(s): af1173e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -31
README.md CHANGED
@@ -122,34 +122,4 @@ print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"],
122
 
123
  ## Training
124
 
125
- The Common Voice `train` and `validation` datasets were used for training. The script used for training can be found [here](https://github.com/jqueguiner/wav2vec2-sprint/blob/main/run_common_voice.py).
126
- The parameters passed were:
127
-
128
- ```bash
129
- #!/usr/bin/env bash
130
- python run_common_voice.py \\
131
- --model_name_or_path="facebook/wav2vec2-large-xlsr-53" \\
132
- --dataset_config_name="pt" \\
133
- --output_dir=/workspace/output_models/pt/wav2vec2-large-xlsr-pt \\
134
- --cache_dir=/workspace/data \\
135
- --overwrite_output_dir \\
136
- --num_train_epochs="30" \\
137
- --per_device_train_batch_size="32" \\
138
- --per_device_eval_batch_size="32" \\
139
- --evaluation_strategy="steps" \\
140
- --learning_rate="3e-4" \\
141
- --warmup_steps="500" \\
142
- --fp16 \\
143
- --freeze_feature_extractor \\
144
- --save_steps="500" \\
145
- --eval_steps="500" \\
146
- --save_total_limit="1" \\
147
- --logging_steps="500" \\
148
- --group_by_length \\
149
- --feat_proj_dropout="0.0" \\
150
- --layerdrop="0.1" \\
151
- --gradient_checkpointing \\
152
- --do_train --do_eval \\
153
- ```
154
-
155
- Notebook containing the evaluation can be found [here](https://colab.research.google.com/drive/1pejk9gv9vMcUOjyVQ_vsV2ngW4NiWLWy?usp=sharing).
 
122
 
123
  ## Training
124
 
125
+ The Common Voice `train` and `validation` datasets were used for training. The script used for training can be found [here](https://colab.research.google.com/drive/1pejk9gv9vMcUOjyVQ_vsV2ngW4NiWLWy?usp=sharing).