| base_model: unsloth/Phi-3-mini-4k-instruct | |
| config: | |
| dpo: | |
| beta: 0.1 | |
| max_length: 1024 | |
| lora: | |
| alpha: 24 | |
| dropout: 0.1 | |
| r: 12 | |
| target_modules: | |
| - q_proj | |
| - v_proj | |
| - k_proj | |
| - o_proj | |
| model: | |
| base_model: unsloth/Phi-3-mini-4k-instruct | |
| quantization_config: 4bit | |
| use_quantization: true | |
| training: | |
| batch_size: 2 | |
| eval_steps: 300 | |
| gradient_accumulation_steps: 4 | |
| learning_rate: 5e-6 | |
| logging_steps: 10 | |
| num_epochs: 3 | |
| save_steps: 300 | |
| warmup_steps: 75 | |
| merged_model_dir: outputs/models/dpo_phi3_20250721_060254/merged_model | |
| model_info: | |
| config_file: configs/models/phi3.yaml | |
| memory_requirement: medium | |
| name: Phi-3 Mini | |
| size: 3.8B | |
| training_time: medium | |
| model_key: phi3 | |
| output_dir: outputs/models/dpo_phi3_20250721_060254 | |
| training_completed: '2025-07-21T06:11:21.508248' | |