lmind_nq_train6000_eval6489_v1_qa_Qwen_Qwen1.5-4B_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_nq_train6000_eval6489_v1_qa dataset. It achieves the following results on the evaluation set:

• Loss: 2.4726
• Accuracy: 0.5579

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: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 20.0

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
1.7657	0.9973	187	0.5738	1.6215
1.497	2.0	375	0.5742	1.6180
1.2345	2.9973	562	0.5713	1.6951
1.0084	4.0	750	0.5659	1.8059
0.8397	4.9973	937	0.5647	1.9245
0.7186	6.0	1125	0.5614	2.0345
0.6421	6.9973	1312	0.5608	2.1148
0.5968	8.0	1500	0.5585	2.1779
0.5417	8.9973	1687	0.5568	2.2654
0.5356	9.9733	1870	0.5594	2.2527
0.5261	10.9973	2057	2.3376	0.5585
0.5179	12.0	2245	2.3704	0.5595
0.5116	12.9973	2432	2.3617	0.5589
0.5056	14.0	2620	2.4022	0.5581
0.5063	14.9973	2807	2.3861	0.5587
0.4796	16.0	2995	2.3658	0.5585
0.4757	16.9973	3182	2.4195	0.5577
0.4779	18.0	3370	2.4573	0.5573
0.4782	18.9973	3557	2.4896	0.5589
0.4784	19.9733	3740	2.4726	0.5579

Framework versions

• PEFT 0.5.0
• Transformers 4.40.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1