lmind_nq_train6000_eval6489_v1_qa_Qwen_Qwen1.5-4B_3e-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: 2.3067
• Accuracy: 0.5581

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
1.7168	0.9973	187	1.6089	0.5754
1.3336	2.0	375	1.6442	0.5728
0.9813	2.9973	562	1.7657	0.5690
0.7483	4.0	750	1.9240	0.566
0.6395	4.9973	937	2.0308	0.5644
0.5836	6.0	1125	2.0914	0.5626
0.5559	6.9973	1312	2.1673	0.5617
0.5386	8.0	1500	2.1641	0.5619
0.5022	8.9973	1687	2.1993	0.5623
0.5035	10.0	1875	2.2047	0.5633
0.5013	10.9973	2062	2.2971	0.5616
0.5063	12.0	2250	2.2050	0.5618
0.5048	12.9973	2437	2.2624	0.5597
0.506	14.0	2625	2.3161	0.5598
0.511	14.9973	2812	2.2551	0.5554
0.5163	16.0	3000	2.3024	0.5578
0.4861	16.9973	3187	2.2554	0.5585
0.4925	18.0	3375	2.2402	0.5579
0.4927	18.9973	3562	2.2989	0.5570
0.4868	19.9467	3740	2.3067	0.5581

Framework versions

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