lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_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_reciteonly_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 2.7806
• Accuracy: 0.5793

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
1.8147	0.9973	187	1.7073	0.6045
1.6968	2.0	375	1.6898	0.6062
1.6305	2.9973	562	1.6912	0.6071
1.5527	4.0	750	1.7046	0.6062
1.4642	4.9973	937	1.7309	0.6055
1.3868	6.0	1125	1.7706	0.6037
1.278	6.9973	1312	1.8113	0.6016
1.1816	8.0	1500	1.8514	0.6004
1.0781	8.9973	1687	1.9348	0.5973
1.0042	10.0	1875	1.9619	0.5969
0.9175	10.9973	2062	2.0437	0.5938
0.8259	12.0	2250	2.1340	0.5910
0.7665	12.9973	2437	2.1983	0.5899
0.6822	14.0	2625	2.2949	0.5864
0.6189	14.9973	2812	2.3464	0.5854
0.5588	16.0	3000	2.4582	0.5835
0.4806	16.9973	3187	2.5345	0.5829
0.4272	18.0	3375	2.6303	0.5805
0.3805	18.9973	3562	2.6586	0.5806
0.3421	19.9467	3740	2.7806	0.5793

Framework versions

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