End of training

README.md

@@ -18,9 +18,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.2801
-- Wer: 84.6009
-- Cer: 70.1608
+- Loss: 0.5546
+- Wer: 82.4230
+- Cer: 64.5239
 
 ## Model description
 
@@ -39,23 +39,59 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 8e-06
+- learning_rate: 1e-05
 - train_batch_size: 16
 - eval_batch_size: 16
 - seed: 42
 - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: cosine
-- lr_scheduler_warmup_steps: 8000
-- num_epochs: 5
+- lr_scheduler_warmup_steps: 2000
+- num_epochs: 15
 
 ### Training results
 
-| Training Loss | Epoch  | Step  | Validation Loss | Wer     | Cer     |
-|:-------------:|:------:|:-----:|:---------------:|:-------:|:-------:|
-| 0.2359        | 1.2472 | 5000  | 0.2733          | 94.9788 | 93.0093 |
-| 0.1824        | 2.4944 | 10000 | 0.2674          | 66.9301 | 53.0174 |
-| 0.1267        | 3.7416 | 15000 | 0.2587          | 79.5997 | 65.3546 |
-| 0.0775        | 4.9888 | 20000 | 0.2801          | 84.6009 | 70.1608 |
+| Training Loss | Epoch   | Step  | Validation Loss | Wer      | Cer     |
+|:-------------:|:-------:|:-----:|:---------------:|:--------:|:-------:|
+| 0.2956        | 0.3742  | 1500  | 0.3018          | 63.6348  | 47.5770 |
+| 0.2867        | 0.7483  | 3000  | 0.2841          | 93.2416  | 79.6830 |
+| 0.213         | 1.1225  | 4500  | 0.2759          | 97.8988  | 85.8519 |
+| 0.227         | 1.4966  | 6000  | 0.2677          | 91.6266  | 74.8936 |
+| 0.2212        | 1.8708  | 7500  | 0.2617          | 93.4321  | 79.6175 |
+| 0.1611        | 2.2449  | 9000  | 0.2667          | 96.8099  | 78.9492 |
+| 0.1589        | 2.6191  | 10500 | 0.2643          | 100.5175 | 76.5414 |
+| 0.1697        | 2.9933  | 12000 | 0.2622          | 84.5441  | 67.4836 |
+| 0.1071        | 3.3674  | 13500 | 0.2855          | 96.4829  | 81.3859 |
+| 0.1136        | 3.7416  | 15000 | 0.2814          | 99.5195  | 75.9088 |
+| 0.0637        | 4.1157  | 16500 | 0.3122          | 83.2163  | 64.5333 |
+| 0.0692        | 4.4899  | 18000 | 0.3137          | 90.7196  | 76.6609 |
+| 0.0687        | 4.8641  | 19500 | 0.3179          | 91.4560  | 75.3464 |
+| 0.0333        | 5.2382  | 21000 | 0.3506          | 90.0088  | 71.7452 |
+| 0.0366        | 5.6124  | 22500 | 0.3571          | 86.4889  | 69.5759 |
+| 0.0379        | 5.9865  | 24000 | 0.3554          | 88.3228  | 63.6308 |
+| 0.0174        | 6.3607  | 25500 | 0.3917          | 91.0494  | 75.0314 |
+| 0.0176        | 6.7348  | 27000 | 0.3937          | 84.2740  | 64.2078 |
+| 0.0075        | 7.1090  | 28500 | 0.4261          | 83.2419  | 65.5333 |
+| 0.0092        | 7.4832  | 30000 | 0.4228          | 89.9832  | 69.3133 |
+| 0.0096        | 7.8573  | 31500 | 0.4328          | 86.2415  | 71.8841 |
+| 0.0042        | 8.2315  | 33000 | 0.4482          | 83.1168  | 64.2068 |
+| 0.0045        | 8.6056  | 34500 | 0.4502          | 86.0169  | 67.4873 |
+| 0.005         | 8.9798  | 36000 | 0.4580          | 83.4096  | 65.3331 |
+| 0.0027        | 9.3540  | 37500 | 0.4667          | 82.2837  | 60.9284 |
+| 0.0027        | 9.7281  | 39000 | 0.4718          | 84.5469  | 63.7325 |
+| 0.0012        | 10.1023 | 40500 | 0.4819          | 80.8678  | 62.0516 |
+| 0.0012        | 10.4764 | 42000 | 0.4911          | 81.6951  | 59.3660 |
+| 0.0013        | 10.8506 | 43500 | 0.4950          | 78.6415  | 60.5107 |
+| 0.0005        | 11.2247 | 45000 | 0.5076          | 81.8117  | 62.4285 |
+| 0.0005        | 11.5989 | 46500 | 0.5055          | 78.8746  | 59.2439 |
+| 0.0008        | 11.9731 | 48000 | 0.5101          | 79.2244  | 59.3010 |
+| 0.0006        | 12.3472 | 49500 | 0.5234          | 81.2459  | 63.5160 |
+| 0.0002        | 12.7214 | 51000 | 0.5310          | 83.3272  | 65.2183 |
+| 0.0001        | 13.0955 | 52500 | 0.5380          | 83.5347  | 65.1864 |
+| 0.0001        | 13.4697 | 54000 | 0.5450          | 81.4990  | 63.0972 |
+| 0.0001        | 13.8439 | 55500 | 0.5491          | 82.4230  | 64.9317 |
+| 0.0001        | 14.2180 | 57000 | 0.5523          | 82.3889  | 64.7414 |
+| 0.0001        | 14.5922 | 58500 | 0.5544          | 82.5510  | 64.7718 |
+| 0.0001        | 14.9663 | 60000 | 0.5546          | 82.4230  | 64.5239 |
 
 
 ### Framework versions

generation_config.json

@@ -41,10 +41,7 @@
       6
     ]
   ],
-  "begin_suppress_tokens": [
-    220,
-    50257
-  ],
+  "begin_suppress_tokens": [],
   "bos_token_id": 50257,
   "decoder_start_token_id": 50258,
   "eos_token_id": 50257,
@@ -151,102 +148,13 @@
     "<|yue|>": 50358,
     "<|zh|>": 50260
   },
+  "language": "ar",
   "max_initial_timestamp_index": 50,
   "max_length": 448,
   "no_timestamps_token_id": 50364,
   "pad_token_id": 50257,
   "prev_sot_token_id": 50362,
   "return_timestamps": false,
-  "suppress_tokens": [
-    1,
-    2,
-    7,
-    8,
-    9,
-    10,
-    14,
-    25,
-    26,
-    27,
-    28,
-    29,
-    31,
-    58,
-    59,
-    60,
-    61,
-    62,
-    63,
-    90,
-    91,
-    92,
-    93,
-    359,
-    503,
-    522,
-    542,
-    873,
-    893,
-    902,
-    918,
-    922,
-    931,
-    1350,
-    1853,
-    1982,
-    2460,
-    2627,
-    3246,
-    3253,
-    3268,
-    3536,
-    3846,
-    3961,
-    4183,
-    4667,
-    6585,
-    6647,
-    7273,
-    9061,
-    9383,
-    10428,
-    10929,
-    11938,
-    12033,
-    12331,
-    12562,
-    13793,
-    14157,
-    14635,
-    15265,
-    15618,
-    16553,
-    16604,
-    18362,
-    18956,
-    20075,
-    21675,
-    22520,
-    26130,
-    26161,
-    26435,
-    28279,
-    29464,
-    31650,
-    32302,
-    32470,
-    36865,
-    42863,
-    47425,
-    49870,
-    50254,
-    50258,
-    50359,
-    50360,
-    50361,
-    50362,
-    50363
-  ],
   "task": "transcribe",
   "task_to_id": {
     "transcribe": 50360,