| { | |
| "model_id": "0xvoid0000/zira-researcher", | |
| "json_dir": "/root/CRIS-reason-input-initial/json_files", | |
| "num_shards": 5, | |
| "passes_over_all_shards": 1, | |
| "model_context_limit": 65536, | |
| "train_max_seq_len": 8192, | |
| "lora_r": 32, | |
| "lora_alpha": 64, | |
| "lora_dropout": 0.05, | |
| "learning_rate": 0.0001, | |
| "per_device_batch_size": 1, | |
| "grad_accum": 8, | |
| "warmup_steps": 10, | |
| "world_size": 4 | |
| } |