patrickvonplaten
commited on
Commit
·
5da3530
1
Parent(s):
cefa808
Create run.sh
Browse files
run.sh
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env bash
|
2 |
+
python -m torch.distributed.launch \
|
3 |
+
--nproc_per_node 2 run_speech_recognition_ctc.py \
|
4 |
+
--dataset_name="common_voice" \
|
5 |
+
--model_name_or_path="./wav2vec2-xls-r-phoneme-300m-tr" \
|
6 |
+
--tokenizer_name_or_path="./wav2vec2-xls-r-phoneme-300m-tr" \
|
7 |
+
--dataset_config_name="tr" \
|
8 |
+
--phoneme_language="tr" \
|
9 |
+
--output_dir="./test_phoneme_tr" \
|
10 |
+
--overwrite_output_dir \
|
11 |
+
--num_train_epochs="3" \
|
12 |
+
--per_device_train_batch_size="1" \
|
13 |
+
--gradient_accumulation_steps="8" \
|
14 |
+
--learning_rate="5e-5" \
|
15 |
+
--warmup_steps="500" \
|
16 |
+
--evaluation_strategy="steps" \
|
17 |
+
--text_column_name="sentence" \
|
18 |
+
--pad_token="<pad>" \
|
19 |
+
--unk_token="<unk>" \
|
20 |
+
--save_steps="100" \
|
21 |
+
--eval_steps="100" \
|
22 |
+
--logging_steps="1" \
|
23 |
+
--layerdrop="0.0" \
|
24 |
+
--save_total_limit="3" \
|
25 |
+
--freeze_feature_extractor \
|
26 |
+
--chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
|
27 |
+
--fp16 \
|
28 |
+
--group_by_length \
|
29 |
+
--do_eval \
|
30 |
+
--gradient_checkpointing \
|
31 |
+
--use_auth_token \
|
32 |
+
--do_train --do_eval \
|
33 |
+
--push_to_hub \
|