lmind_nq_train6000_eval6489_v1_reciteonly_qa_v3_Qwen_Qwen1.5-4B_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: 2.4062
• Accuracy: 0.5899

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: 0.0001
• 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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.7823	0.9973	187	1.6938	0.6057
1.6543	2.0	375	1.6800	0.6076
1.526	2.9973	562	1.6967	0.6070
1.3835	4.0	750	1.7289	0.6067
1.2248	4.9973	937	1.7874	0.6050
1.0689	6.0	1125	1.8847	0.6018
0.8895	6.9973	1312	1.9975	0.5989
0.7219	8.0	1500	2.1065	0.5960
0.5575	8.9973	1687	2.2722	0.5920
0.4466	9.9733	1870	2.4062	0.5899

Framework versions

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