Model save

README.md

@@ -1,6 +1,6 @@
 ---
 license: apache-2.0
-base_model: facebook/wav2vec2-large-960h
+base_model: facebook/wav2vec2-base-960h
 tags:
 - generated_from_trainer
 metrics:
@@ -15,10 +15,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # Helldivers2ASR_V3
 
-This model is a fine-tuned version of [facebook/wav2vec2-large-960h](https://huggingface.co/facebook/wav2vec2-large-960h) on an unknown dataset.
+This model is a fine-tuned version of [facebook/wav2vec2-base-960h](https://huggingface.co/facebook/wav2vec2-base-960h) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 100.2962
-- Wer: 0.2096
+- Loss: 57.7348
+- Wer: 0.0332
 
 ## Model description
 
@@ -37,9 +37,9 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 0.0005
-- train_batch_size: 16
-- eval_batch_size: 16
+- learning_rate: 0.0002
+- train_batch_size: 32
+- eval_batch_size: 32
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: constant
@@ -50,36 +50,36 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch | Step | Validation Loss | Wer    |
 |:-------------:|:-----:|:----:|:---------------:|:------:|
-| 1044.7378     | 1.0   | 328  | 674.2581        | 0.7134 |
-| 788.3286      | 2.0   | 656  | 551.7861        | 0.5443 |
-| 686.6791      | 3.0   | 984  | 475.5942        | 0.4852 |
-| 630.9951      | 4.0   | 1312 | 385.9394        | 0.4615 |
-| 591.5513      | 5.0   | 1640 | 375.8221        | 0.4019 |
-| 536.7542      | 6.0   | 1968 | 359.7422        | 0.4190 |
-| 503.6451      | 7.0   | 2296 | 351.6354        | 0.3961 |
-| 469.5973      | 8.0   | 2624 | 303.6820        | 0.3580 |
-| 437.5308      | 9.0   | 2952 | 242.7078        | 0.3969 |
-| 422.3093      | 10.0  | 3280 | 271.1854        | 0.3764 |
-| 409.4765      | 11.0  | 3608 | 281.5080        | 0.3541 |
-| 384.7355      | 12.0  | 3936 | 232.6271        | 0.3331 |
-| 368.876       | 13.0  | 4264 | 205.0227        | 0.3607 |
-| 359.576       | 14.0  | 4592 | 206.7654        | 0.3628 |
-| 347.9149      | 15.0  | 4920 | 203.3771        | 0.3131 |
-| 324.0592      | 16.0  | 5248 | 195.5896        | 0.3134 |
-| 313.765       | 17.0  | 5576 | 187.8961        | 0.2963 |
-| 308.2892      | 18.0  | 5904 | 175.5116        | 0.2926 |
-| 298.7906      | 19.0  | 6232 | 204.8492        | 0.2947 |
-| 276.8503      | 20.0  | 6560 | 174.5410        | 0.2763 |
-| 273.6627      | 21.0  | 6888 | 169.8062        | 0.2813 |
-| 262.0091      | 22.0  | 7216 | 151.0183        | 0.2493 |
-| 256.9988      | 23.0  | 7544 | 146.9712        | 0.2522 |
-| 253.0775      | 24.0  | 7872 | 138.4501        | 0.2474 |
-| 250.4051      | 25.0  | 8200 | 107.8654        | 0.2112 |
-| 247.1768      | 26.0  | 8528 | 113.1651        | 0.2062 |
-| 230.0392      | 27.0  | 8856 | 138.3462        | 0.2564 |
-| 232.4302      | 28.0  | 9184 | 128.4431        | 0.2469 |
-| 225.1965      | 29.0  | 9512 | 108.4411        | 0.2661 |
-| 220.6905      | 30.0  | 9840 | 100.2962        | 0.2096 |
+| 1440.5784     | 1.0   | 165  | 827.7584        | 0.3982 |
+| 901.7305      | 2.0   | 330  | 516.9960        | 0.2632 |
+| 680.9522      | 3.0   | 495  | 461.0711        | 0.2190 |
+| 572.9876      | 4.0   | 660  | 373.8802        | 0.1709 |
+| 455.9722      | 5.0   | 825  | 325.0266        | 0.1536 |
+| 405.2042      | 6.0   | 990  | 315.5705        | 0.1397 |
+| 335.4539      | 7.0   | 1155 | 229.3944        | 0.1115 |
+| 311.7148      | 8.0   | 1320 | 184.8958        | 0.0918 |
+| 292.6152      | 9.0   | 1485 | 177.3645        | 0.0840 |
+| 268.9287      | 10.0  | 1650 | 150.6056        | 0.0748 |
+| 224.9027      | 11.0  | 1815 | 134.2426        | 0.0746 |
+| 221.6842      | 12.0  | 1980 | 152.2036        | 0.0633 |
+| 216.7037      | 13.0  | 2145 | 94.9223         | 0.0510 |
+| 163.8511      | 14.0  | 2310 | 93.6272         | 0.0395 |
+| 169.5703      | 15.0  | 2475 | 124.6715        | 0.0568 |
+| 149.0854      | 16.0  | 2640 | 133.8979        | 0.0607 |
+| 157.8008      | 17.0  | 2805 | 66.3784         | 0.0330 |
+| 147.502       | 18.0  | 2970 | 71.3790         | 0.0335 |
+| 148.8542      | 19.0  | 3135 | 72.7481         | 0.0419 |
+| 136.0251      | 20.0  | 3300 | 129.4439        | 0.0523 |
+| 129.7949      | 21.0  | 3465 | 49.1545         | 0.0262 |
+| 126.8121      | 22.0  | 3630 | 67.8833         | 0.0343 |
+| 126.8445      | 23.0  | 3795 | 66.8705         | 0.0327 |
+| 127.7805      | 24.0  | 3960 | 70.8930         | 0.0309 |
+| 113.9002      | 25.0  | 4125 | 44.5920         | 0.0233 |
+| 102.9785      | 26.0  | 4290 | 65.2669         | 0.0280 |
+| 107.9157      | 27.0  | 4455 | 39.7528         | 0.0212 |
+| 103.8186      | 28.0  | 4620 | 46.9734         | 0.0246 |
+| 103.4841      | 29.0  | 4785 | 58.0863         | 0.0264 |
+| 81.7682       | 30.0  | 4950 | 57.7348         | 0.0332 |
 
 
 ### Framework versions

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/wav2vec2-large-960h",
+  "_name_or_path": "facebook/wav2vec2-base-960h",
   "activation_dropout": 0.1,
   "adapter_attn_dim": null,
   "adapter_kernel_size": 3,
@@ -53,12 +53,13 @@
   "feat_proj_dropout": 0.1,
   "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.15,
+  "gradient_checkpointing": false,
   "hidden_act": "gelu",
   "hidden_dropout": 0.15,
   "hidden_dropout_prob": 0.1,
-  "hidden_size": 1024,
+  "hidden_size": 768,
   "initializer_range": 0.02,
-  "intermediate_size": 4096,
+  "intermediate_size": 3072,
   "layer_norm_eps": 1e-05,
   "layerdrop": 0.1,
   "mask_feature_length": 10,
@@ -69,15 +70,15 @@
   "mask_time_prob": 0.05,
   "model_type": "wav2vec2",
   "num_adapter_layers": 3,
-  "num_attention_heads": 16,
+  "num_attention_heads": 12,
   "num_codevector_groups": 2,
   "num_codevectors_per_group": 320,
   "num_conv_pos_embedding_groups": 16,
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
-  "num_hidden_layers": 24,
+  "num_hidden_layers": 12,
   "num_negatives": 100,
-  "output_hidden_size": 1024,
+  "output_hidden_size": 768,
   "pad_token_id": 0,
   "proj_codevector_dim": 256,
   "tdnn_dilation": [

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf39c3bad3692b8ada5e3dbea6704a07b1768bafeeb113bc0fc2160f0feba24a
-size 1261897368
+oid sha256:d81fe311e050fc59cbc06a44537263e8399d7b27e2915852a9aeb51171a1e58b
+size 377611120

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c6e5da415c451c4bb69a5677f24ce26f9e584b30080767c1bfc04bf9fccee97
+oid sha256:837b93a71322b18603c527653dcae8bb68872145d8929ffaa0f6430370f6b1c6
 size 5176