lmind_nq_train6000_eval6489_v1_qa_Qwen_Qwen1.5-4B_5e-5_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_nq_train6000_eval6489_v1_qa dataset. It achieves the following results on the evaluation set:

• Loss: 2.5214
• Accuracy: 0.5534

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: 5e-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
1.8353	0.9973	187	1.6327	0.5719
1.5773	2.0	375	1.6119	0.5743
1.4151	2.9973	562	1.6409	0.5734
1.226	4.0	750	1.7002	0.5688
1.0718	4.9973	937	1.7919	0.5664
0.9478	6.0	1125	1.8953	0.5631
0.8356	6.9973	1312	1.9827	0.5607
0.7482	8.0	1500	2.0659	0.5591
0.6415	8.9973	1687	2.2042	0.5565
0.6094	10.0	1875	2.2516	0.5552
0.5807	10.9973	2062	2.2925	0.5554
0.5647	12.0	2250	2.3210	0.5562
0.5523	12.9973	2437	2.3624	0.556
0.5417	14.0	2625	2.4357	0.5541
0.5335	14.9973	2812	2.4554	0.5528
0.5312	16.0	3000	2.4443	0.5553
0.5001	16.9973	3187	2.4833	0.5533
0.4992	18.0	3375	2.4588	0.5559
0.5004	18.9973	3562	2.5082	0.5546
0.4983	19.9467	3740	2.5214	0.5534

Framework versions

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