| { |
| "started_at_utc": "2026-03-19T11:54:34.606942+00:00", |
| "ended_at_utc": "2026-03-19T13:30:48.781371+00:00", |
| "train_elapsed_sec": 5774.174, |
| "stopped_by_time_limit": false, |
| "global_step": 43, |
| "train_metrics": { |
| "train_runtime": 5773.5801, |
| "train_samples_per_second": 0.118, |
| "train_steps_per_second": 0.007, |
| "total_flos": 2.771601971163955e+16, |
| "train_loss": 0.9663696178170138, |
| "epoch": 1.0 |
| }, |
| "args": { |
| "model_name": "vodkaslime/codellama-7b-hf", |
| "train_file": "/kaggle/working/python-llm-grpo-new-techniques/training/data/train.jsonl", |
| "val_file": "/kaggle/working/python-llm-grpo-new-techniques/training/data/val.jsonl", |
| "output_dir": "/kaggle/working/python-llm-grpo-new-techniques/training/output/phase2-codellama-7b-lora-kaggle-1h", |
| "block_size": 1024, |
| "num_train_epochs": 1, |
| "max_steps": -1, |
| "max_runtime_minutes": 55.0, |
| "max_runtime_seconds": 7200.0, |
| "effective_max_runtime_seconds": 7200.0, |
| "per_device_train_batch_size": 1, |
| "per_device_eval_batch_size": 1, |
| "gradient_accumulation_steps": 16, |
| "learning_rate": 0.0002, |
| "save_steps": 100, |
| "eval_steps": 100, |
| "save_total_limit": 2, |
| "warmup_steps": 50, |
| "use_4bit": true, |
| "bf16": true, |
| "fp16": false, |
| "gradient_checkpointing": true, |
| "seed": 42, |
| "artifact_mode": "adapter_only" |
| } |
| } |
