lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_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: 3.9177
• Accuracy: 0.4908

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: 0.0001
• 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	Accuracy	Validation Loss
2.2624	1.0	250	0.5159	2.3220
2.0942	2.0	500	0.5176	2.3289
1.8479	3.0	750	0.5148	2.3997
1.6153	4.0	1000	0.5107	2.5067
1.3618	5.0	1250	0.5052	2.6641
1.1477	6.0	1500	0.5016	2.8411
0.9248	7.0	1750	0.4978	3.0246
0.7705	8.0	2000	0.4954	3.2090
0.6344	9.0	2250	0.4935	3.3400
0.5612	10.0	2500	0.4926	3.4933
0.4967	11.0	2750	3.5794	0.4917
0.4696	12.0	3000	3.6326	0.4914
0.4399	13.0	3250	3.7408	0.4920
0.4324	14.0	3500	3.7450	0.4915
0.4105	15.0	3750	3.8301	0.4922
0.4081	16.0	4000	3.8488	0.4921
0.3939	17.0	4250	3.8492	0.4913
0.3924	18.0	4500	3.8751	0.4915
0.382	19.0	4750	3.9337	0.4910
0.3832	20.0	5000	3.9177	0.4908

Framework versions

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