lmind_nq_train6000_eval6489_v1_doc_qa_v3_Qwen_Qwen1.5-4B_5e-5_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_nq_train6000_eval6489_v1_doc_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 2.5129
• Accuracy: 0.5609

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-05
• 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	Validation Loss	Accuracy
1.843	1.0	529	1.6193	0.5738
1.7946	2.0	1058	1.6033	0.5756
1.675	3.0	1587	1.6412	0.5741
1.6045	4.0	2116	1.7177	0.5717
1.4962	5.0	2645	1.8518	0.5677
1.4058	6.0	3174	1.9380	0.5668
1.348	7.0	3703	2.0013	0.5651
1.3061	8.0	4232	2.1522	0.5629
1.2537	9.0	4761	2.1630	0.5635
1.208	10.0	5290	2.2419	0.5630
1.1346	11.0	5819	2.2893	0.5629
1.0951	12.0	6348	2.3280	0.5639
1.0306	13.0	6877	2.3896	0.5619
0.9875	14.0	7406	2.3976	0.5630
0.9005	15.0	7935	2.4602	0.5610
0.8609	16.0	8464	2.4857	0.5624
0.7923	17.0	8993	2.4992	0.5619
0.7522	18.0	9522	2.5021	0.5613
0.7038	19.0	10051	2.5203	0.5623
0.6484	20.0	10580	2.5129	0.5609

Framework versions

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