This model is a fine-tuned version of Qwen/Qwen3-8B on the Final_selected_5k_cleaned_hermes_replaced dataset. It achieves the following results on the evaluation set:

Loss: 0.3307

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-06
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 8
total_eval_batch_size: 4
optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.3019	0.1761	100	0.3737
0.2678	0.3521	200	0.3432
0.331	0.5282	300	0.3353
0.2923	0.7042	400	0.3274
0.3057	0.8803	500	0.3131
0.2293	1.0563	600	0.3185
0.2162	1.2324	700	0.3176
0.21	1.4085	800	0.3175
0.1958	1.5845	900	0.3163
0.2206	1.7606	1000	0.3114
0.229	1.9366	1100	0.3109
0.1808	2.1127	1200	0.3253
0.1908	2.2887	1300	0.3278
0.1952	2.4648	1400	0.3295
0.2336	2.6408	1500	0.3302
0.1763	2.8169	1600	0.3302
0.1618	2.9930	1700	0.3290

Framework versions

Transformers 4.55.0
Pytorch 2.8.0+cu128
Datasets 3.6.0
Tokenizers 0.21.1

lemonhat/Qwen3-8B-Final_selected_5k_cleaned_hermes_replaced

Final_selected_5k_cleaned_hermes_replaced

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

lemonhat/
Qwen3-8B-Final_selected_5k_cleaned_hermes_replaced