learning_rate: 5e-06
train_batch_size: 1
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 2
total_train_batch_size: 16
total_eval_batch_size: 32
optimizer: Use adamw_torch 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: 0.05
num_epochs: 3.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss
0.0968	0.0746	5	0.0690
0.0346	0.1493	10	0.0366
0.0256	0.2239	15	0.0321
0.0328	0.2985	20	0.0302
0.0287	0.3731	25	0.0280
0.0221	0.4478	30	0.0275
0.0347	0.5224	35	0.0265
0.0212	0.5970	40	0.0255
0.0245	0.6716	45	0.0254
0.0209	0.7463	50	0.0244
0.0242	0.8209	55	0.0239
0.0226	0.8955	60	0.0236
0.0208	0.9701	65	0.0232
0.0135	1.0448	70	0.0229
0.0141	1.1194	75	0.0234
0.0165	1.1940	80	0.0235
0.0173	1.2687	85	0.0233
0.0123	1.3433	90	0.0231
0.0145	1.4179	95	0.0232
0.0154	1.4925	100	0.0226
0.0147	1.5672	105	0.0224
0.0132	1.6418	110	0.0228
0.0155	1.7164	115	0.0227
0.0149	1.7910	120	0.0221
0.0169	1.8657	125	0.0219
0.0136	1.9403	130	0.0218
0.0139	2.0149	135	0.0219
0.0101	2.0896	140	0.0220
0.0087	2.1642	145	0.0222
0.0089	2.2388	150	0.0223
0.0112	2.3134	155	0.0225
0.0083	2.3881	160	0.0227
0.008	2.4627	165	0.0227
0.0109	2.5373	170	0.0228
0.0103	2.6119	175	0.0228
0.008	2.6866	180	0.0229
0.0103	2.7612	185	0.0229
0.008	2.8358	190	0.0229
0.0109	2.9104	195	0.0229
0.009	2.9851	200	0.0229

wuyue0129/LLM4RecAug

LLM_Rec_Qwen2.5_7B_full_sft