test
Browse files- {checkpoint-10000 β first-save}/added_tokens.json +0 -0
- {checkpoint-10000 β first-save}/config.json +0 -0
- {checkpoint-10000 β first-save}/flax_model.msgpack +0 -0
- {checkpoint-10000 β first-save}/generation_config.json +0 -0
- {checkpoint-10000 β first-save}/merges.txt +0 -0
- {checkpoint-10000 β first-save}/preprocessor_config.json +0 -0
- {checkpoint-10000 β first-save}/special_tokens_map.json +0 -0
- {checkpoint-10000 β first-save}/tokenizer_config.json +0 -0
- {checkpoint-10000 β first-save}/train_state.msgpack +0 -0
- {checkpoint-10000 β first-save}/vocab.json +0 -0
- run_experiment2.sh +5 -5
{checkpoint-10000 β first-save}/added_tokens.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/config.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/flax_model.msgpack
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/generation_config.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/merges.txt
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/preprocessor_config.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/special_tokens_map.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/tokenizer_config.json
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/train_state.msgpack
RENAMED
File without changes
|
{checkpoint-10000 β first-save}/vocab.json
RENAMED
File without changes
|
run_experiment2.sh
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
#!/usr/bin/env bash
|
2 |
TOKENIZERS_PARALLELISM=false python3 run_distillation.py \
|
3 |
-
--model_name_or_path "./
|
4 |
--teacher_model_name_or_path "NbAiLab/nb-whisper-large" \
|
5 |
--train_dataset_name "NbAiLab/annotated_ncc_speech_styling_v2_vad3_distil_postLv2" \
|
6 |
--train_dataset_config_name "" \
|
@@ -10,16 +10,16 @@ TOKENIZERS_PARALLELISM=false python3 run_distillation.py \
|
|
10 |
--eval_split_name "validation_norwegian_fleurs" \
|
11 |
--eval_steps 500 \
|
12 |
--save_steps 5000 \
|
13 |
-
--warmup_steps
|
14 |
--learning_rate 0.0003 \
|
15 |
--lr_scheduler_type "constant_with_warmup" \
|
16 |
--logging_steps 500 \
|
17 |
-
--save_total_limit
|
18 |
--max_steps 200000 \
|
19 |
--wer_threshold 10 \
|
20 |
--per_device_train_batch_size 4\
|
21 |
--per_device_eval_batch_size 4 \
|
22 |
-
--dataloader_num_workers
|
23 |
--dtype "bfloat16" \
|
24 |
--output_dir "./nb-distil-whisper-large-flax7" \
|
25 |
--do_train \
|
@@ -33,7 +33,7 @@ TOKENIZERS_PARALLELISM=false python3 run_distillation.py \
|
|
33 |
--use_auth_token \
|
34 |
--report_to "wandb" \
|
35 |
--wandb_project "nb-distil-whisper-large-fleurseval" \
|
36 |
-
--wandb_name "
|
37 |
--save_code_to_wandb \
|
38 |
--save_train_state \
|
39 |
--hub_model_id "NbAiLab/nb-distil-whisper-large-flax7"\
|
|
|
1 |
#!/usr/bin/env bash
|
2 |
TOKENIZERS_PARALLELISM=false python3 run_distillation.py \
|
3 |
+
--model_name_or_path "./first-save" \
|
4 |
--teacher_model_name_or_path "NbAiLab/nb-whisper-large" \
|
5 |
--train_dataset_name "NbAiLab/annotated_ncc_speech_styling_v2_vad3_distil_postLv2" \
|
6 |
--train_dataset_config_name "" \
|
|
|
10 |
--eval_split_name "validation_norwegian_fleurs" \
|
11 |
--eval_steps 500 \
|
12 |
--save_steps 5000 \
|
13 |
+
--warmup_steps 0 \
|
14 |
--learning_rate 0.0003 \
|
15 |
--lr_scheduler_type "constant_with_warmup" \
|
16 |
--logging_steps 500 \
|
17 |
+
--save_total_limit 2 \
|
18 |
--max_steps 200000 \
|
19 |
--wer_threshold 10 \
|
20 |
--per_device_train_batch_size 4\
|
21 |
--per_device_eval_batch_size 4 \
|
22 |
+
--dataloader_num_workers 32 \
|
23 |
--dtype "bfloat16" \
|
24 |
--output_dir "./nb-distil-whisper-large-flax7" \
|
25 |
--do_train \
|
|
|
33 |
--use_auth_token \
|
34 |
--report_to "wandb" \
|
35 |
--wandb_project "nb-distil-whisper-large-fleurseval" \
|
36 |
+
--wandb_name "flax_experiment2_bs4_v5_1e4_wer10_restart10k" \
|
37 |
--save_code_to_wandb \
|
38 |
--save_train_state \
|
39 |
--hub_model_id "NbAiLab/nb-distil-whisper-large-flax7"\
|