lmind_nq_train6000_eval6489_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_nq_train6000_eval6489_v1_qa dataset. It achieves the following results on the evaluation set:

• Loss: 2.4467
• Accuracy: 0.5472

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
1.7152	0.9973	187	1.6235	0.5736
1.2597	2.0	375	1.6922	0.5689
0.9105	2.9973	562	1.8390	0.5645
0.7214	4.0	750	1.9820	0.5616
0.6428	4.9973	937	2.0143	0.56
0.5948	6.0	1125	2.1345	0.5575
0.5692	6.9973	1312	2.1957	0.5575
0.5569	8.0	1500	2.2088	0.5549
0.5176	8.9973	1687	2.2513	0.5565
0.5272	10.0	1875	2.2161	0.5550
0.5305	10.9973	2062	2.2532	0.5537
0.5463	12.0	2250	2.2262	0.5543
0.5471	12.9973	2437	2.2971	0.5516
0.5436	14.0	2625	2.2834	0.5515
0.5417	14.9973	2812	2.3678	0.5468
0.5409	16.0	3000	2.3382	0.5494
0.4996	16.9973	3187	2.4009	0.5493
0.4995	18.0	3375	2.4317	0.5487
0.5015	18.9973	3562	2.4855	0.5474
0.4986	19.9467	3740	2.4467	0.5472

Framework versions

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