finalform/rewrite_files_Qwen3-Coder-30B-A3B-Instruct

rewrite_results

This model is a fine-tuned version of Qwen/Qwen3-Coder-30B-A3B-Instruct on the train dataset. It achieves the following results on the evaluation set:

• Loss: 0.0616
• Accuracy: 0.9850

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: 0.0004
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 8
• 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: 0.085
• num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.1481	2	0.2477	0.9611
0.2441	0.2963	4	0.1997	0.9648
0.2367	0.4444	6	0.1587	0.9688
0.2367	0.5926	8	0.1476	0.9710
0.1895	0.7407	10	0.1318	0.9732
0.1361	0.8889	12	0.1172	0.9759
0.1361	1.0	14	0.1053	0.9783
0.18	1.1481	16	0.0985	0.9792
0.1193	1.2963	18	0.0932	0.9798
0.1193	1.4444	20	0.0875	0.9804
0.0823	1.5926	22	0.0840	0.9806
0.1175	1.7407	24	0.0778	0.9814
0.1175	1.8889	26	0.0737	0.9827
0.0898	2.0	28	0.0704	0.9836
0.0948	2.1481	30	0.0689	0.9838
0.0948	2.2963	32	0.0670	0.9841
0.0739	2.4444	34	0.0653	0.9841
0.0526	2.5926	36	0.0643	0.9845
0.0526	2.7407	38	0.0634	0.9845
0.0564	2.8889	40	0.0625	0.9847
0.0678	3.0	42	0.0616	0.9850
0.0678	3.1481	44	0.0615	0.9852
0.0499	3.2963	46	0.0616	0.9853
0.0437	3.4444	48	0.0620	0.9853
0.0437	3.5926	50	0.0621	0.9851
0.0421	3.7407	52	0.0624	0.9852
0.0557	3.8889	54	0.0624	0.9852
0.0557	4.0	56	0.0623	0.9851

Framework versions

• PEFT 0.17.1
• Transformers 4.57.1
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2