Qwen2.5-Coder-1.5B-Instruct-Airscript

This model is a fine-tuned version of Qwen/Qwen2.5-Coder-1.5B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4578

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 30
• training_steps: 1600

Training results

Training Loss	Epoch	Step	Validation Loss
2.1783	0.0625	100	2.1006
1.8426	0.125	200	1.8535
1.7343	0.1875	300	1.7350
1.6313	0.25	400	1.6520
1.5817	0.3125	500	1.5982
1.5498	0.375	600	1.5604
1.5019	0.4375	700	1.5322
1.4852	0.5	800	1.5103
1.461	0.5625	900	1.4939
1.4483	0.625	1000	1.4820
1.4434	0.6875	1100	1.4723
1.4254	0.75	1200	1.4659
1.4224	0.8125	1300	1.4619
1.4188	0.875	1400	1.4596
1.4245	0.9375	1500	1.4585
1.4172	1.0	1600	1.4578

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.5.0
• Datasets 3.0.1
• Tokenizers 0.20.1