| # Fine-tuned Model: Dindan | |
| ## 📚 Training Configuration | |
| - **data_path**: `QomSSLab/MMPD_QA_v2` | |
| - **output_dir**: `gemma312b_lora_chckpnts-dindan` | |
| - **new_model_name**: `Dindan` | |
| - **data_ratio**: `1.0` | |
| - **model_name**: `QomSSLab/gemma-3-12b-it` | |
| - **use_4bit**: `False` | |
| - **use_lora**: `False` | |
| - **max_seq_length**: `2048` | |
| - **batch_size**: `16` | |
| - **gradient_accu**: `8` | |
| - **epochs**: `1` | |
| - **learning_rate**: `2e-05` | |
| - **lora_alpha**: `256` | |
| - **lora_drop**: `0.05` | |
| - **lora_r**: `256` | |
| - **tune_embedding_layer**: `False` | |
| - **hf_token**: `********` | |
| - **resume_from_checkpoint**: `False` | |
| - **use_8bit_optimizer**: `True` | |
| - **push_to_hub**: `True` | |
| - **push_lora_only**: `False` | |
| - **train_only_on_assistant**: `False` | |
| - **last_response**: `False` | |
| --- | |
| Auto-generated after training. | |