constantinedivis/whisper-tiny-rus

README.md

@@ -4,26 +4,11 @@ license: apache-2.0
 base_model: openai/whisper-tiny
 tags:
 - generated_from_trainer
-datasets:
-- common_voice_17_0
 metrics:
 - wer
 model-index:
 - name: whisper-tiny-rus
-  results:
-  - task:
-      name: Automatic Speech Recognition
-      type: automatic-speech-recognition
-    dataset:
-      name: common_voice_17_0
-      type: common_voice_17_0
-      config: ru
-      split: test
-      args: ru
-    metrics:
-    - name: Wer
-      type: wer
-      value: 39.12070231891154
+  results: []
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -31,11 +16,10 @@ should probably proofread and complete it, then remove this comment. -->
 
 # whisper-tiny-rus
 
-This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the common_voice_17_0 dataset.
+This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.5293
-- Wer Ortho: 44.5430
-- Wer: 39.1207
+- Loss: 0.4508
+- Wer: 37.6577
 
 ## Model description
 
@@ -55,25 +39,30 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 1e-05
-- train_batch_size: 16
-- eval_batch_size: 16
+- train_batch_size: 128
+- eval_batch_size: 128
 - 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: constant_with_warmup
-- lr_scheduler_warmup_steps: 50
-- training_steps: 500
+- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 100
+- training_steps: 600
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss | Wer Ortho | Wer     |
-|:-------------:|:------:|:----:|:---------------:|:---------:|:-------:|
-| 0.301         | 0.3032 | 500  | 0.5293          | 44.5430   | 39.1207 |
+| Training Loss | Epoch  | Step | Validation Loss | Wer     |
+|:-------------:|:------:|:----:|:---------------:|:-------:|
+| 0.6166        | 0.3984 | 100  | 0.6163          | 45.3709 |
+| 0.5109        | 0.7968 | 200  | 0.5225          | 41.1251 |
+| 0.4615        | 1.1952 | 300  | 0.4850          | 39.6391 |
+| 0.4377        | 1.5936 | 400  | 0.4664          | 38.5069 |
+| 0.433         | 1.9920 | 500  | 0.4544          | 37.6695 |
+| 0.4186        | 2.3904 | 600  | 0.4508          | 37.6577 |
 
 
 ### Framework versions
 
-- Transformers 4.48.3
-- Pytorch 2.5.1+cu124
-- Datasets 3.2.0
-- Tokenizers 0.21.0
+- Transformers 4.49.0
+- Pytorch 2.1.0+cu118
+- Datasets 3.4.0
+- Tokenizers 0.21.1

generation_config.json

@@ -32,16 +32,6 @@
   "bos_token_id": 50257,
   "decoder_start_token_id": 50258,
   "eos_token_id": 50257,
-  "forced_decoder_ids": [
-    [
-      1,
-      null
-    ],
-    [
-      2,
-      50359
-    ]
-  ],
   "is_multilingual": true,
   "lang_to_id": {
     "<|af|>": 50327,
@@ -244,5 +234,5 @@
     "transcribe": 50359,
     "translate": 50358
   },
-  "transformers_version": "4.48.3"
+  "transformers_version": "4.49.0"
 }