lmind_nq_train6000_eval6489_v1_doc_qa_v3_Qwen_Qwen1.5-4B_5e-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: 2.2417
• Accuracy: 0.5610

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: 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
1.8583	1.0	529	1.6376	0.5726
1.6329	2.0	1058	1.6881	0.5713
1.3464	3.0	1587	1.8256	0.5663
1.1624	4.0	2116	1.9223	0.5652
0.964	5.0	2645	1.9720	0.5643
0.8117	6.0	3174	2.0016	0.5647
0.7242	7.0	3703	2.0785	0.5639
0.6381	8.0	4232	2.0954	0.5645
0.573	9.0	4761	2.1067	0.5623
0.5269	10.0	5290	2.1356	0.5646
0.5144	11.0	5819	2.1951	0.5616
0.4887	12.0	6348	2.1779	0.5631
0.4636	13.0	6877	2.1757	0.5611
0.467	14.0	7406	2.1781	0.5624
0.4613	15.0	7935	2.2312	0.5612
0.4405	16.0	8464	2.1800	0.5629
0.4308	17.0	8993	2.1960	0.5628
0.4401	18.0	9522	2.2355	0.5610
0.4334	19.0	10051	2.2380	0.5608
0.4218	20.0	10580	2.2417	0.5610

Framework versions

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