lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_3e-5_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: 3.6297
• 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: 3e-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
2.2869	1.0	250	2.3483	0.5139
2.208	2.0	500	2.3308	0.5160
2.1226	3.0	750	2.3344	0.5159
2.0165	4.0	1000	2.3550	0.5151
1.8949	5.0	1250	2.4099	0.5125
1.7858	6.0	1500	2.4598	0.5108
1.6743	7.0	1750	2.5374	0.5080
1.578	8.0	2000	2.6112	0.5064
1.4609	9.0	2250	2.6753	0.5041
1.3821	10.0	2500	2.7747	0.5018
1.2732	11.0	2750	2.8579	0.4979
1.2047	12.0	3000	2.9205	0.4972
1.0965	13.0	3250	3.0633	0.4941
1.0197	14.0	3500	3.1028	0.4933
0.9397	15.0	3750	3.2329	0.4913
0.8754	16.0	4000	3.3337	0.4890
0.8084	17.0	4250	3.4384	0.4885
0.7655	18.0	4500	3.4688	0.4875
0.7125	19.0	4750	3.5650	0.4876
0.6824	20.0	5000	3.6297	0.4864

Framework versions

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