lmind_hotpot_train8000_eval7405_v1_doc_qa_Qwen_Qwen1.5-4B_3e-4_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_hotpot_train8000_eval7405_v1_doc_qa dataset. It achieves the following results on the evaluation set:

• Loss: 3.1631
• Accuracy: 0.5165

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
1.7567	0.9998	1089	2.2626	0.5190
1.5617	1.9995	2178	2.2436	0.5246
1.343	2.9993	3267	2.3385	0.5237
1.1682	4.0	4357	2.4995	0.5215
1.0141	4.9998	5446	2.6397	0.5182
0.9023	5.9995	6535	2.7929	0.5170
0.8008	6.9993	7624	2.8233	0.5162
0.7377	8.0	8714	2.8833	0.5180
0.6732	8.9998	9803	2.9550	0.5165
0.6225	9.9995	10892	2.9767	0.5165
0.5858	10.9993	11981	3.0117	0.5165
0.5618	12.0	13071	3.0317	0.5170
0.5464	12.9998	14160	3.0686	0.5167
0.5243	13.9995	15249	3.0829	0.5149
0.5066	14.9993	16338	3.0958	0.5127
0.4947	16.0	17428	3.0921	0.5153
0.4841	16.9998	18517	3.1170	0.5162
0.4727	17.9995	19606	3.1375	0.5172
0.4634	18.9993	20695	3.1323	0.5150
0.4468	19.9954	21780	3.1631	0.5165

Framework versions

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