decompiler-distil-best

This model is a fine-tuned version of Qwen/Qwen3-4B-Thinking-2507 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6512

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_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: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
2.647	0.0752	50	0.4902
2.1412	0.1503	100	0.4627
1.9836	0.2255	150	0.4513
2.2814	0.3006	200	0.4546
1.8827	0.3758	250	0.4733
1.8347	0.4510	300	0.5015
2.4491	0.5261	350	0.5301
1.9778	0.6013	400	0.5414
1.9626	0.6764	450	0.5532
2.0233	0.7516	500	0.5841
2.075	0.8268	550	0.6085
1.5818	0.9019	600	0.6284
2.033	0.9771	650	0.6512

Framework versions

• PEFT 0.18.0
• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1