48 lines
1.4 KiB
JSON
48 lines
1.4 KiB
JSON
|
{
|
||
|
|
"experiment_metadata": {
|
||
|
|
"timestamp": "2026-08-04T15:02:33.496432+00:00",
|
||
|
|
"pretrained_model_name_or_path": "unsloth/Qwen3-4B-unsloth-bnb-4bit",
|
||
|
|
"data_dir": "nityaak/stance-matrix-mtcsd-ezstance",
|
||
|
|
"push_to_hub_id": "nityaak/qwen3-4b-stance-matrix-mtcsd-ezstance-qlora"
|
||
|
|
},
|
||
|
|
"configuration": {
|
||
|
|
"load_in_4bit": true,
|
||
|
|
"from_foundation_model": false,
|
||
|
|
"learning_rate": 2e-05,
|
||
|
|
"num_train_epochs": 1.0,
|
||
|
|
"per_device_train_batch_size": 8,
|
||
|
|
"gradient_accumulation_steps": 2,
|
||
|
|
"effective_batch_size": 16,
|
||
|
|
"weight_decay": 0.01,
|
||
|
|
"warmup_ratio": 0.1,
|
||
|
|
"optim": "adamw_8bit",
|
||
|
|
"max_length": 512,
|
||
|
|
"packing": true,
|
||
|
|
"completion_only_loss": true
|
||
|
|
},
|
||
|
|
"training_stats": {
|
||
|
|
"train_runtime": 1493.2053,
|
||
|
|
"train_samples_per_second": 2.08,
|
||
|
|
"train_steps_per_second": 0.131,
|
||
|
|
"total_flos": 3.194777643575808e+16,
|
||
|
|
"train_loss": 0.32319403871511804,
|
||
|
|
"epoch": 1.0
|
||
|
|
},
|
||
|
|
"gpu_stats": {
|
||
|
|
"gpu_id": 0,
|
||
|
|
"gpu_name": "NVIDIA A100-SXM4-40GB",
|
||
|
|
"memory_used_mb": 14822.0,
|
||
|
|
"memory_total_mb": 40626.8,
|
||
|
|
"memory_utilization_percent": 36.48,
|
||
|
|
"runtime_seconds": 1493.2
|
||
|
|
},
|
||
|
|
"weight_verification": {
|
||
|
|
"all_changed": true,
|
||
|
|
"unchanged_params": []
|
||
|
|
},
|
||
|
|
"carbon_tracking": {
|
||
|
|
"enabled": true,
|
||
|
|
"log_dir": "carbontracker",
|
||
|
|
"note": "See carbontracker/ in this repo for actual consumption (time, energy, CO2eq)."
|
||
|
|
}
|
||
|
|
}
|