ihanif commited on
Commit
8a2a133
·
1 Parent(s): 982fe5b

Training in progress, step 5100

Browse files
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5cf9831b7679721474b5e26b01f9539d2048ee48467b5ab2326bd974bf7beb0d
3
  size 967102601
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b62a476b62cb2d81e03a8841574b0a1adbb552e5d2728c9b09122585cd304dd0
3
  size 967102601
run-2.sh CHANGED
@@ -6,18 +6,18 @@ python run_speech_recognition_seq2seq_streaming.py \
6
  --train_split_name="train+validation" \
7
  --eval_split_name="test" \
8
  --model_index_name="Whisper Small Pashto" \
9
- --max_steps="5000" \
10
  --output_dir="./" \
11
  --per_device_train_batch_size="32" \
12
  --per_device_eval_batch_size="16" \
13
  --gradient_accumulation_steps="2" \
14
  --logging_steps="25" \
15
  --learning_rate="3e-7" \
16
- --warmup_steps="500" \
17
  --evaluation_strategy="steps" \
18
- --eval_steps="500" \
19
  --save_strategy="steps" \
20
- --save_steps="500" \
21
  --generation_max_length="225" \
22
  --length_column_name="input_length" \
23
  --max_duration_in_seconds="30" \
@@ -25,6 +25,8 @@ python run_speech_recognition_seq2seq_streaming.py \
25
  --freeze_feature_encoder="False" \
26
  --report_to="tensorboard" \
27
  --metric_for_best_model="wer" \
 
 
28
  --gradient_checkpointing \
29
  --fp16 \
30
  --overwrite_output_dir="False" \
 
6
  --train_split_name="train+validation" \
7
  --eval_split_name="test" \
8
  --model_index_name="Whisper Small Pashto" \
9
+ --max_steps="5200" \
10
  --output_dir="./" \
11
  --per_device_train_batch_size="32" \
12
  --per_device_eval_batch_size="16" \
13
  --gradient_accumulation_steps="2" \
14
  --logging_steps="25" \
15
  --learning_rate="3e-7" \
16
+ --warmup_steps="10" \
17
  --evaluation_strategy="steps" \
18
+ --eval_steps="100" \
19
  --save_strategy="steps" \
20
+ --save_steps="100" \
21
  --generation_max_length="225" \
22
  --length_column_name="input_length" \
23
  --max_duration_in_seconds="30" \
 
25
  --freeze_feature_encoder="False" \
26
  --report_to="tensorboard" \
27
  --metric_for_best_model="wer" \
28
+ --greater_is_better="False" \
29
+ --load_best_model_at_end \
30
  --gradient_checkpointing \
31
  --fp16 \
32
  --overwrite_output_dir="False" \
runs/Dec10_08-38-21_132-145-129-157/1670661513.746379/events.out.tfevents.1670661513.132-145-129-157.104887.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57030f5cec50e2169055381e07762bf8ab96778346fa65e53c0ba9a1ba157d44
3
+ size 5871
runs/Dec10_08-38-21_132-145-129-157/events.out.tfevents.1670661513.132-145-129-157.104887.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d111d00242f409c0dc26aab3733dc4457b3aa0ffcb4b0d4a374c3563773a0dc
3
+ size 5212
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2d15d72002cecf8323f6cd941cec892fce6b19d6e69aa1365a60139bd64d32ff
3
  size 3579
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe3ea30c17ecb01004af4af0c24a4b66bb08abdb557a33104926cc836852de7a
3
  size 3579