Fix RNN-T tokenizer: <blank> at id 1024 (aliased as pad), consistent with blank_token_id=1024
Browse files- config.json +1 -1
- processor_config.json +1 -0
- tokenizer.json +0 -9
- tokenizer_config.json +1 -1
config.json
CHANGED
|
@@ -5,7 +5,6 @@
|
|
| 5 |
"blank_token_id": 1024,
|
| 6 |
"decoder_hidden_size": 640,
|
| 7 |
"dtype": "float32",
|
| 8 |
-
"durations": [],
|
| 9 |
"encoder_config": {
|
| 10 |
"activation_dropout": 0.1,
|
| 11 |
"attention_bias": true,
|
|
@@ -38,6 +37,7 @@
|
|
| 38 |
"model_type": "parakeet_rnnt",
|
| 39 |
"num_decoder_layers": 2,
|
| 40 |
"pad_token_id": 0,
|
|
|
|
| 41 |
"transformers_version": "5.10.0.dev0",
|
| 42 |
"vocab_size": 1025
|
| 43 |
}
|
|
|
|
| 5 |
"blank_token_id": 1024,
|
| 6 |
"decoder_hidden_size": 640,
|
| 7 |
"dtype": "float32",
|
|
|
|
| 8 |
"encoder_config": {
|
| 9 |
"activation_dropout": 0.1,
|
| 10 |
"attention_bias": true,
|
|
|
|
| 37 |
"model_type": "parakeet_rnnt",
|
| 38 |
"num_decoder_layers": 2,
|
| 39 |
"pad_token_id": 0,
|
| 40 |
+
"rnnt_loss_reduction": "mean_volume",
|
| 41 |
"transformers_version": "5.10.0.dev0",
|
| 42 |
"vocab_size": 1025
|
| 43 |
}
|
processor_config.json
CHANGED
|
@@ -1,5 +1,6 @@
|
|
| 1 |
{
|
| 2 |
"blank_token": "<blank>",
|
|
|
|
| 3 |
"feature_extractor": {
|
| 4 |
"feature_extractor_type": "ParakeetFeatureExtractor",
|
| 5 |
"feature_size": 80,
|
|
|
|
| 1 |
{
|
| 2 |
"blank_token": "<blank>",
|
| 3 |
+
"decoder_type": "rnnt",
|
| 4 |
"feature_extractor": {
|
| 5 |
"feature_extractor_type": "ParakeetFeatureExtractor",
|
| 6 |
"feature_size": 80,
|
tokenizer.json
CHANGED
|
@@ -14,15 +14,6 @@
|
|
| 14 |
},
|
| 15 |
{
|
| 16 |
"id": 1024,
|
| 17 |
-
"content": "<pad>",
|
| 18 |
-
"single_word": false,
|
| 19 |
-
"lstrip": false,
|
| 20 |
-
"rstrip": false,
|
| 21 |
-
"normalized": false,
|
| 22 |
-
"special": true
|
| 23 |
-
},
|
| 24 |
-
{
|
| 25 |
-
"id": 1025,
|
| 26 |
"content": "<blank>",
|
| 27 |
"single_word": false,
|
| 28 |
"lstrip": false,
|
|
|
|
| 14 |
},
|
| 15 |
{
|
| 16 |
"id": 1024,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
"content": "<blank>",
|
| 18 |
"single_word": false,
|
| 19 |
"lstrip": false,
|
tokenizer_config.json
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
"backend": "tokenizers",
|
| 3 |
"clean_up_tokenization_spaces": false,
|
| 4 |
"model_max_length": 1000000000000000019884624838656,
|
| 5 |
-
"pad_token": "<
|
| 6 |
"processor_class": "ParakeetProcessor",
|
| 7 |
"tokenizer_class": "ParakeetTokenizer",
|
| 8 |
"unk_token": "<unk>"
|
|
|
|
| 2 |
"backend": "tokenizers",
|
| 3 |
"clean_up_tokenization_spaces": false,
|
| 4 |
"model_max_length": 1000000000000000019884624838656,
|
| 5 |
+
"pad_token": "<blank>",
|
| 6 |
"processor_class": "ParakeetProcessor",
|
| 7 |
"tokenizer_class": "ParakeetTokenizer",
|
| 8 |
"unk_token": "<unk>"
|