learning_rate: 1e-05
train_batch_size: 1
eval_batch_size: 1
seed: 3407
gradient_accumulation_steps: 16
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_8BIT 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
lr_scheduler_warmup_steps: 4
num_epochs: 3

Training Loss	Epoch	Step	Validation Loss
0.7617	0.2486	23	0.9320
0.8866	0.4973	46	0.7884
0.8458	0.7459	69	0.7398
0.8002	0.9946	92	0.7191
0.742	1.2378	115	0.7069
0.645	1.4865	138	0.6987
0.5373	1.7351	161	0.6927
0.9976	1.9838	184	0.6890
0.8863	2.2270	207	0.6866
0.8198	2.4757	230	0.6851
0.5288	2.7243	253	0.6847
0.8511	2.9730	276	0.6845

gutsartificial/Qwen3-8B-2025-07-30_18-19-12

Qwen3-8B-2025-07-30_18-19-12