Hugging Face's logo

esc-bench
/
wav2vec2-aed-common_voice

Automatic Speech Recognition
Transformers
JAX
English
speech-encoder-decoder
esb
Model card Files
xet
 Community
Sanchit Gandhi commited on
Commit
66c2e56
·
1 Parent(s): b4d031f

Correct scripts

Browse files
Files changed (2) hide show
  1. README.md +6 -4
  2. run_common_voice.sh +6 -4
README.md CHANGED
@@ -12,7 +12,7 @@ To reproduce this run, execute:
12
  python run_flax_speech_recognition_seq2seq.py \
13
  --dataset_name="esc-benchmark/esc-datasets" \
14
  --model_name_or_path="esc-benchmark/wav2vec2-aed-pretrained" \
15
- --dataset_config_name="ami" \
16
  --output_dir="./" \
17
  --wandb_name="wav2vec2-aed-common-voice" \
18
  --wandb_project="wav2vec2-aed" \
@@ -21,11 +21,13 @@ python run_flax_speech_recognition_seq2seq.py \
21
  --learning_rate="1e-4" \
22
  --warmup_steps="500" \
23
  --logging_steps="25" \
24
- --max_steps="50000" \
25
  --eval_steps="10000" \
26
  --save_steps="10000" \
27
- --generation_max_length="200" \
28
- --generation_num_beams="5" \
 
 
29
  --generation_length_penalty="1.2" \
30
  --max_eval_duration_in_seconds="20" \
31
  --overwrite_output_dir \
 
12
  python run_flax_speech_recognition_seq2seq.py \
13
  --dataset_name="esc-benchmark/esc-datasets" \
14
  --model_name_or_path="esc-benchmark/wav2vec2-aed-pretrained" \
15
+ --dataset_config_name="common_voice" \
16
  --output_dir="./" \
17
  --wandb_name="wav2vec2-aed-common-voice" \
18
  --wandb_project="wav2vec2-aed" \
 
21
  --learning_rate="1e-4" \
22
  --warmup_steps="500" \
23
  --logging_steps="25" \
24
+ --max_steps="50001" \
25
  --eval_steps="10000" \
26
  --save_steps="10000" \
27
+ --generation_max_length="40" \
28
+ --generation_num_beams="1" \
29
+ --final_generation_max_length="200" \
30
+ --generation_num_beams="14" \
31
  --generation_length_penalty="1.2" \
32
  --max_eval_duration_in_seconds="20" \
33
  --overwrite_output_dir \
run_common_voice.sh CHANGED
@@ -2,7 +2,7 @@
2
  python run_flax_speech_recognition_seq2seq.py \
3
  --dataset_name="esc-benchmark/esc-datasets" \
4
  --model_name_or_path="esc-benchmark/wav2vec2-aed-pretrained" \
5
- --dataset_config_name="ami" \
6
  --output_dir="./" \
7
  --wandb_name="wav2vec2-aed-common-voice" \
8
  --wandb_project="wav2vec2-aed" \
@@ -11,11 +11,13 @@ python run_flax_speech_recognition_seq2seq.py \
11
  --learning_rate="1e-4" \
12
  --warmup_steps="500" \
13
  --logging_steps="25" \
14
- --max_steps="50000" \
15
  --eval_steps="10000" \
16
  --save_steps="10000" \
17
- --generation_max_length="200" \
18
- --generation_num_beams="5" \
 
 
19
  --generation_length_penalty="1.2" \
20
  --max_eval_duration_in_seconds="20" \
21
  --overwrite_output_dir \
 
2
  python run_flax_speech_recognition_seq2seq.py \
3
  --dataset_name="esc-benchmark/esc-datasets" \
4
  --model_name_or_path="esc-benchmark/wav2vec2-aed-pretrained" \
5
+ --dataset_config_name="common_voice" \
6
  --output_dir="./" \
7
  --wandb_name="wav2vec2-aed-common-voice" \
8
  --wandb_project="wav2vec2-aed" \
 
11
  --learning_rate="1e-4" \
12
  --warmup_steps="500" \
13
  --logging_steps="25" \
14
+ --max_steps="50001" \
15
  --eval_steps="10000" \
16
  --save_steps="10000" \
17
+ --generation_max_length="40" \
18
+ --generation_num_beams="1" \
19
+ --final_generation_max_length="200" \
20
+ --generation_num_beams="14" \
21
  --generation_length_penalty="1.2" \
22
  --max_eval_duration_in_seconds="20" \
23
  --overwrite_output_dir \