lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_5e-4_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_hotpot_train8000_eval7405_v1_qa dataset. It achieves the following results on the evaluation set:

• Loss: 3.7344
• Accuracy: 0.4864

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
2.252	1.0	250	2.3165	0.5171
1.8363	2.0	500	2.4264	0.5127
1.3801	3.0	750	2.6120	0.5059
1.0246	4.0	1000	2.8617	0.5008
0.7286	5.0	1250	3.0953	0.4959
0.601	6.0	1500	3.2139	0.4950
0.5138	7.0	1750	3.2912	0.4933
0.4837	8.0	2000	3.4517	0.49
0.4506	9.0	2250	3.4107	0.4911
0.4578	10.0	2500	3.4786	0.4905
0.4362	11.0	2750	3.5410	0.4899
0.4429	12.0	3000	3.5656	0.4909
0.4366	13.0	3250	3.5425	0.4890
0.4474	14.0	3500	3.5998	0.4900
0.4283	15.0	3750	3.6044	0.4870
0.4299	16.0	4000	3.6720	0.4882
0.4202	17.0	4250	3.6220	0.4860
0.4318	18.0	4500	3.6682	0.4875
0.4151	19.0	4750	3.7105	0.4857
0.4227	20.0	5000	3.7344	0.4864

Framework versions

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