Commit
•
d1558b4
1
Parent(s):
c4e5ff9
Saving weights and logs of epoch 0
Browse files- config.json +2 -0
- create_model.py +1 -1
- flax_model.msgpack +1 -1
- run_librispeech.sh +2 -0
- tokenizer_config.json +1 -1
config.json
CHANGED
@@ -1,7 +1,9 @@
|
|
1 |
{
|
|
|
2 |
"architectures": [
|
3 |
"SpeechEncoderDecoderModel"
|
4 |
],
|
|
|
5 |
"decoder": {
|
6 |
"_name_or_path": "facebook/bart-large-cnn",
|
7 |
"_num_labels": 3,
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "./",
|
3 |
"architectures": [
|
4 |
"SpeechEncoderDecoderModel"
|
5 |
],
|
6 |
+
"cache_dir": "/home/sanchit_huggingface_co/cache/huggingface/transformers",
|
7 |
"decoder": {
|
8 |
"_name_or_path": "facebook/bart-large-cnn",
|
9 |
"_num_labels": 3,
|
create_model.py
CHANGED
@@ -4,7 +4,7 @@ from transformers import AutoFeatureExtractor, AutoTokenizer, FlaxSpeechEncoderD
|
|
4 |
encoder_id = "facebook/wav2vec2-large-lv60"
|
5 |
decoder_id = "facebook/bart-large-cnn"
|
6 |
|
7 |
-
model = FlaxSpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_id, decoder_id, encoder_add_adapter=True)
|
8 |
|
9 |
model.config.encoder.feat_proj_dropout = 0.0
|
10 |
model.config.encoder.final_dropout = 0.0
|
|
|
4 |
encoder_id = "facebook/wav2vec2-large-lv60"
|
5 |
decoder_id = "facebook/bart-large-cnn"
|
6 |
|
7 |
+
model = FlaxSpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_id, decoder_id, encoder_add_adapter=True, cache_dir="/home/sanchit_huggingface_co/cache/huggingface/transormers")
|
8 |
|
9 |
model.config.encoder.feat_proj_dropout = 0.0
|
10 |
model.config.encoder.final_dropout = 0.0
|
flax_model.msgpack
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 2353635949
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:932441fd4d1167a741bd188a1f0bc953a229e065d97f15027e824bb6dbe4ae7b
|
3 |
size 2353635949
|
run_librispeech.sh
CHANGED
@@ -6,6 +6,8 @@ python run_flax_speech_recognition_seq2seq.py \
|
|
6 |
--train_split_name="train.100" \
|
7 |
--eval_split_name="validation" \
|
8 |
--output_dir="./" \
|
|
|
|
|
9 |
--preprocessing_num_workers="16" \
|
10 |
--length_column_name="input_length" \
|
11 |
--overwrite_output_dir \
|
|
|
6 |
--train_split_name="train.100" \
|
7 |
--eval_split_name="validation" \
|
8 |
--output_dir="./" \
|
9 |
+
--dataset_cache_dir="/home/sanchit_huggingface_co/cache/huggingface/datasets" \
|
10 |
+
--cache_dir="/home/sanchit_huggingface_co/cache/huggingface/transformers" \
|
11 |
--preprocessing_num_workers="16" \
|
12 |
--length_column_name="input_length" \
|
13 |
--overwrite_output_dir \
|
tokenizer_config.json
CHANGED
@@ -1 +1 @@
|
|
1 |
-
{"errors": "replace", "bos_token": "<s>", "eos_token": "</s>", "sep_token": "</s>", "cls_token": "<s>", "unk_token": "<unk>", "pad_token": "<pad>", "mask_token": "<mask>", "add_prefix_space": false, "trim_offsets": true, "model_max_length": 1024, "special_tokens_map_file": null, "name_or_path": "
|
|
|
1 |
+
{"errors": "replace", "bos_token": "<s>", "eos_token": "</s>", "sep_token": "</s>", "cls_token": "<s>", "unk_token": "<unk>", "pad_token": "<pad>", "mask_token": "<mask>", "add_prefix_space": false, "trim_offsets": true, "model_max_length": 1024, "special_tokens_map_file": null, "name_or_path": "./", "tokenizer_class": "BartTokenizer"}
|