jdrt_byclass_rinnna_hubert_asr_1

README.md

@@ -0,0 +1,95 @@
+---
+license: apache-2.0
+base_model: rinna/japanese-hubert-base
+tags:
+- generated_from_trainer
+metrics:
+- wer
+model-index:
+- name: jdrt_byclass_rinnna_hubert_asr_1
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# jdrt_byclass_rinnna_hubert_asr_1
+
+This model is a fine-tuned version of [rinna/japanese-hubert-base](https://huggingface.co/rinna/japanese-hubert-base) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.3647
+- Wer: 0.4190
+- Cer: 0.2827
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 256
+- eval_batch_size: 256
+- seed: 42
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 250
+- num_epochs: 50
+
+### Training results
+
+| Training Loss | Epoch | Step | Validation Loss | Wer    | Cer    |
+|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
+| 10.589        | 1.0   | 53   | 5.6588          | 0.9156 | 0.9495 |
+| 5.0974        | 2.0   | 106  | 4.0914          | 0.9156 | 0.9495 |
+| 3.6701        | 3.0   | 159  | 3.1732          | 0.9156 | 0.9495 |
+| 2.9285        | 4.0   | 212  | 2.7238          | 0.9156 | 0.9495 |
+| 2.6943        | 5.0   | 265  | 2.6600          | 0.9156 | 0.9495 |
+| 2.4567        | 6.0   | 318  | 2.2231          | 0.9960 | 0.9112 |
+| 2.1447        | 7.0   | 371  | 1.9716          | 0.9960 | 0.9112 |
+| 1.8452        | 8.0   | 424  | 1.5058          | 0.9062 | 0.7431 |
+| 1.4358        | 9.0   | 477  | 1.0988          | 0.7370 | 0.5347 |
+| 1.1898        | 10.0  | 530  | 0.9512          | 0.6981 | 0.5062 |
+| 1.0261        | 11.0  | 583  | 0.8354          | 0.6510 | 0.4779 |
+| 0.8371        | 12.0  | 636  | 0.7158          | 0.5560 | 0.3784 |
+| 0.7896        | 13.0  | 689  | 0.6381          | 0.5330 | 0.3686 |
+| 0.6846        | 14.0  | 742  | 0.5720          | 0.5183 | 0.3555 |
+| 0.6357        | 15.0  | 795  | 0.5879          | 0.5030 | 0.3505 |
+| 0.5893        | 16.0  | 848  | 0.5501          | 0.4884 | 0.3468 |
+| 0.558         | 17.0  | 901  | 0.4291          | 0.4487 | 0.3154 |
+| 0.5019        | 18.0  | 954  | 0.4354          | 0.4552 | 0.3064 |
+| 0.4784        | 19.0  | 1007 | 0.4199          | 0.4490 | 0.3014 |
+| 0.4564        | 20.0  | 1060 | 0.4439          | 0.4508 | 0.3153 |
+| 0.4291        | 21.0  | 1113 | 0.4143          | 0.4352 | 0.2845 |
+| 0.4144        | 22.0  | 1166 | 0.4415          | 0.4384 | 0.2812 |
+| 0.3766        | 23.0  | 1219 | 0.3706          | 0.4264 | 0.2918 |
+| 0.3792        | 24.0  | 1272 | 0.3933          | 0.4377 | 0.3015 |
+| 0.3759        | 25.0  | 1325 | 0.3708          | 0.4231 | 0.3023 |
+| 0.337         | 26.0  | 1378 | 0.3762          | 0.4250 | 0.2942 |
+| 0.3282        | 27.0  | 1431 | 0.3595          | 0.4253 | 0.2937 |
+| 0.3174        | 28.0  | 1484 | 0.3998          | 0.4269 | 0.2898 |
+| 0.3156        | 29.0  | 1537 | 0.4056          | 0.4268 | 0.3093 |
+| 0.2921        | 30.0  | 1590 | 0.3694          | 0.4274 | 0.3041 |
+| 0.2929        | 31.0  | 1643 | 0.3917          | 0.4228 | 0.2881 |
+| 0.2686        | 32.0  | 1696 | 0.3880          | 0.4276 | 0.2969 |
+| 0.2776        | 33.0  | 1749 | 0.4038          | 0.4273 | 0.2963 |
+| 0.2619        | 34.0  | 1802 | 0.3647          | 0.4190 | 0.2827 |
+
+
+### Framework versions
+
+- Transformers 4.34.0.dev0
+- Pytorch 2.0.1+cu118
+- Datasets 2.14.5
+- Tokenizers 0.13.3

config.json

@@ -0,0 +1,72 @@
+{
+  "_name_or_path": "rinna/japanese-hubert-base",
+  "activation_dropout": 0.1,
+  "apply_spec_augment": true,
+  "architectures": [
+    "HubertForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "conv_bias": false,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "do_stable_layer_norm": false,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.1,
+  "feat_proj_layer_norm": true,
+  "final_dropout": 0.1,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.05,
+  "model_type": "hubert",
+  "num_attention_heads": 12,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 12,
+  "pad_token_id": 28,
+  "torch_dtype": "float32",
+  "transformers_version": "4.34.0.dev0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 29
+}

preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fbad066105339ed50405ab1b41a4056f95e7c65ecd3981306547fdc2c132bb5
+size 377649057

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:effc66fb5f8dea434e199ed16151b4fa0912bbf14cb4fd4021a07b451e468e03
+size 4027