lmind_nq_train6000_eval6489_v1_doc_qa_v3_Qwen_Qwen1.5-4B_3e-4_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: 2.2532
• Accuracy: 0.5641

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: 0.0003
• 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
1.8369	1.0	529	1.6032	0.5751
1.6451	2.0	1058	1.6357	0.5746
1.3703	3.0	1587	1.7677	0.5716
1.1817	4.0	2116	1.8587	0.5718
0.9674	5.0	2645	1.9319	0.5713
0.7936	6.0	3174	1.9934	0.5704
0.67	7.0	3703	2.0467	0.5684
0.5604	8.0	4232	2.1218	0.5693
0.4747	9.0	4761	2.1342	0.5682
0.4191	10.0	5290	2.1679	0.5674
0.3971	11.0	5819	2.2081	0.5658
0.3753	12.0	6348	2.1840	0.5664
0.3571	13.0	6877	2.2324	0.5634
0.3526	14.0	7406	2.2190	0.5632
0.35	15.0	7935	2.2086	0.5639
0.3323	16.0	8464	2.2655	0.5654
0.3281	17.0	8993	2.2444	0.5667
0.3328	18.0	9522	2.2597	0.5626
0.3305	19.0	10051	2.2682	0.5633
0.3228	20.0	10580	2.2532	0.5641

Framework versions

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