Update README.md

README.md

@@ -1,54 +1,78 @@
 ---
 library_name: transformers
-tags:
-- generated_from_trainer
+datasets:
+- Bingsu/zeroth-korean
+- google/fleurs
+language:
+- ko
+metrics:
+- cer
+- wer
+- bleu
+base_model:
+- microsoft/Phi-4-multimodal-instruct
 model-index:
-- name: Phi-4-multimodal-instruct-ko-speech
-  results: []
+- name: Phi-4-multimodal-instruct-ko-asr
+  results:
+  - task:
+      type: automatic-speech-recognition
+    dataset:
+      type: Bingsu/zeroth_korean
+      name: zeroth-korean-test
+    metrics:
+    - type: bleu
+      name: zeroth-test-BLEU
+      value: 94.837
+    - type: cer
+      name: zeroth-test-CER
+      value: 1.316
+    - type: wer
+      name: zeroth-test-WER
+      value: 2.951
+  - task:
+      type: automatic-speech-recognition
+    dataset:
+      type: google/flerus
+      name: flerus-ko-test
+    metrics:
+    - type: bleu
+      name: fleurs-test-BLEU
+      value: 67.659
+    - type: cer
+      name: fleurs-test-CER
+      value: 7.951
+    - type: wer
+      name: fleurs-test-WER
+      value: 18.313
+pipeline_tag: automatic-speech-recognition
 ---
 
-<!-- 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. -->
 
-# Phi-4-multimodal-instruct-ko-speech
 
-This model was trained from scratch on an unknown dataset.
+This model is fine-tuned from [microsoft/Phi-4-multimodal-instruct](https://huggingface.co/microsoft/Phi-4-multimodal-instruct) on [Bingsu/zeroth-korean](https://huggingface.co/datasets/Bingsu/zeroth-korean), [google/flerus](https://huggingface.co/datasets/Bingsu/google/flerus) in 5 epochs.
 
-## Model description
+This model is trained 960 steps on datasets for Korean Audio Speech Recognition on H100.
 
-More information needed
+After that, we continue training with [CoVoST2 Dataset](https://huggingface.co/datasets/junnei/covost2) / [Only for Korean](https://huggingface.co/datasets/junnei/covost2-ko) for AST.
 
-## Intended uses & limitations
+## Evaluation
 
-More information needed
+Evaluation was done on the following datasets:
+- ASR (Automatic Speech Recognition): Evaluated with CER (Character Error Rate) on zeroth-test set (457 samples).
+- AST (Automatic Speech Translation): Evaluated with BLEU score on fleurs ko <-> en speech translation result (270 samples).
 
-## Training and evaluation data
+Script is retrieved from [here](https://gist.github.com/seastar105/d1d8983b27611370528e3b194dcc5577#file-evaluate-py).
 
-More information needed
+Compared to [Phi-4-mm-inst-zeroth-kor](https://huggingface.co/seastar105/Phi-4-mm-inst-zeroth-kor) and  [Phi-4-multimodal-finetune-ko-speech](https://huggingface.co/daekeun-ml/Phi-4-multimodal-finetune-ko-speech), ASR is significantly improved.
 
-## Training procedure
+| Model                                          | zeroth-CER  | zeroth-WER | fleurs-ko2en | fleurs-ko2en-cot | fleurs-en2ko | fleurs-en2ko-cot |
+|------------------------------------------------|-------------|------------|--------------|------------------|--------------|------------------|
+| original                                       |  198.32     |     -      |     5.63     |       2.42       |     6.86     |       4.17       |
+| daekeun-ml/Phi-4-multimodal-finetune-ko-speech |  1.61       |    3.54    |     7.67     |       8.38       |     12.31    |       9.69       |
+| seastar105/Phi-4-mm-inst-zeroth-kor            |  7.02       |     -      |     7.07     |       9.19       |     13.08    |       9.35       |
+| [**ASR finetune**][ASR]                        |  **1.31**   |    2.95    |     7.46     |       6.24       |     12.15    |       8.91       |
+| + 1 epoch finetune with [Covost-Ko][Covost2-ko]|  3.88       |     -      |   **8.07**   |     **10.09**    |   **18.82**  |    **15.41**     |
+| **AST finetuned model(this model)**            |  **1.77**   |  **2.99**  |   **8.01**   |     **9.09**     |   **17.09**  |    **11.82**     | 
 
-### Training hyperparameters
-
-The following hyperparameters were used during training:
-- learning_rate: 4e-05
-- train_batch_size: 32
-- eval_batch_size: 8
-- seed: 42
-- gradient_accumulation_steps: 4
-- total_train_batch_size: 128
-- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.95) and epsilon=1e-07 and optimizer_args=No additional optimizer arguments
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_steps: 50
-- num_epochs: 2
-
-### Training results
-
-
-
-### Framework versions
-
-- Transformers 4.48.2
-- Pytorch 2.6.0+cu124
-- Datasets 3.3.2
-- Tokenizers 0.21.0
+[Covost2-ko]: https://huggingface.co/datasets/junnei/covost2-ko
+[ASR]: https://huggingface.co/junnei/Phi-4-multimodal-instruct-ko-asr