lmind_nq_train6000_eval6489_v1_qa_Qwen_Qwen1.5-4B_3e-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: 2.5308
• Accuracy: 0.5512

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: 3e-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
1.9148	0.9973	187	1.6426	0.5706
1.6179	2.0	375	1.6183	0.5730
1.5284	2.9973	562	1.6164	0.5749
1.3953	4.0	750	1.6388	0.5725
1.2664	4.9973	937	1.6860	0.5706
1.1544	6.0	1125	1.7575	0.5677
1.0393	6.9973	1312	1.8340	0.5649
0.946	8.0	1500	1.9019	0.5624
0.814	8.9973	1687	2.0181	0.5598
0.7542	10.0	1875	2.0828	0.5576
0.6946	10.9973	2062	2.1505	0.5564
0.6544	12.0	2250	2.2276	0.5562
0.6226	12.9973	2437	2.2688	0.5541
0.5958	14.0	2625	2.3551	0.5526
0.5789	14.9973	2812	2.4366	0.5510
0.5672	16.0	3000	2.4187	0.5533
0.5294	16.9973	3187	2.4533	0.5522
0.5261	18.0	3375	2.4909	0.5514
0.5227	18.9973	3562	2.5120	0.5519
0.5196	19.9467	3740	2.5308	0.5512

Framework versions

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