File size: 1,271 Bytes
c71fb60 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 | {
"experiment_name": "Code-Autopsy QLoRA Bug-Fixing Fine-Tuning v2",
"timestamp": "2026-08-23T04:45:00+05:30",
"base_model": {
"name": "Qwen/Qwen2.5-Coder-7B-Instruct",
"quantization": "4-bit NF4",
"initial_train_loss": 2.162,
"initial_eval_loss": 1.397,
"initial_token_accuracy": "60.29%"
},
"finetuned_model": {
"name": "Code-Autopsy QLoRA Adapter v2",
"training_framework": "PEFT / TRL SFTTrainer",
"final_train_loss": 0.5211,
"final_eval_loss": 0.2442,
"final_token_accuracy": "93.20%",
"best_step_loss": 0.2472
},
"training_metrics": {
"epochs": 3,
"total_steps": 246,
"eval_loss_reduction": "-82.5%",
"accuracy_gain": "+32.91%",
"effective_batch_size": 8,
"learning_rate": 0.0002,
"lr_scheduler": "cosine"
},
"cloud_logging": {
"platform": "Weights & Biases",
"wandb_run_url": "https://wandb.ai/devanshtyagi1903-innothoughts/code-autopsy/runs/gc70q2q2",
"wandb_project_url": "https://wandb.ai/devanshtyagi1903-innothoughts/code-autopsy"
},
"hardware_specs": {
"gpu": "NVIDIA GeForce RTX 5060 (8GB VRAM)",
"quantization": "4-bit NF4 (bitsandbytes)",
"precision": "bfloat16",
"optimizer": "adamw_8bit"
}
} |