| { | |
| "experiment_id": "18_batch8", | |
| "timestamp": "2025-11-19T22:58:37.520255", | |
| "config": { | |
| "experiment_id": "18_batch8", | |
| "base_model": "unsloth/Llama-3.2-3B-Instruct", | |
| "dataset_path": "/home/vincent/wizard101/experiments/guardreasoner/data/guardreasoner_train_chatml.json", | |
| "output_dir": "/home/vincent/wizard101/experiments/guardreasoner/models/exp_18_batch8_lora", | |
| "max_seq_length": 2048, | |
| "load_in_4bit": true, | |
| "lora_r": 16, | |
| "lora_alpha": 16, | |
| "lora_dropout": 0, | |
| "learning_rate": 5e-05, | |
| "num_train_epochs": 3, | |
| "per_device_train_batch_size": 8, | |
| "gradient_accumulation_steps": 12, | |
| "warmup_steps": 100, | |
| "logging_steps": 10, | |
| "save_steps": 250, | |
| "save_total_limit": 3, | |
| "dataloader_num_workers": 4, | |
| "optim": "adamw_8bit" | |
| }, | |
| "training_stats": { | |
| "training_time_hours": 27.97807869858212, | |
| "samples_processed": 382632, | |
| "final_loss": 0.7134220191701062, | |
| "peak_gpu_memory_gb": 11.639, | |
| "vram_percentage": 74.743 | |
| } | |
| } |