learning_rate: 1e-05
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 4
total_train_batch_size: 16
total_eval_batch_size: 4
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3

Training Loss	Epoch	Step	Validation Loss
0.4285	0.2247	100	0.3938
0.3557	0.4494	200	0.3298
0.279	0.6742	300	0.2701
0.2699	0.8989	400	0.2321
0.1477	1.1236	500	0.2274
0.148	1.3483	600	0.2106
0.1349	1.5730	700	0.1972
0.1297	1.7978	800	0.1865
0.0568	2.0225	900	0.1913
0.0494	2.2472	1000	0.2046
0.0459	2.4719	1100	0.2020
0.0509	2.6966	1200	0.2031
0.0495	2.9213	1300	0.2021

Rain-air/Qwen2-7B-osgenesis-plan-sft_0124

sft