lmind_nq_train6000_eval6489_v1_recite_qa_v3_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_recite_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9359
• Accuracy: 0.7135

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
1.86	1.0	529	1.6920	0.6057
1.8271	2.0	1058	1.6426	0.6119
1.7324	3.0	1587	1.5998	0.6163
1.6818	4.0	2116	1.5580	0.6220
1.5864	5.0	2645	1.5211	0.6278
1.5204	6.0	3174	1.4863	0.6327
1.4481	7.0	3703	1.4517	0.6372
1.3768	8.0	4232	1.4121	0.6429
1.2946	9.0	4761	1.3739	0.6482
1.243	10.0	5290	1.3364	0.6532
1.1425	11.0	5819	1.2968	0.6594
1.0847	12.0	6348	1.2539	0.6652
1.0152	13.0	6877	1.2164	0.6706
0.9498	14.0	7406	1.1770	0.6758
0.8652	15.0	7935	1.1323	0.6821
0.8265	16.0	8464	1.0826	0.6901
0.7432	17.0	8993	1.0463	0.6960
0.7106	18.0	9522	1.0098	0.7022
0.669	19.0	10051	0.9696	0.7078
0.6043	20.0	10580	0.9359	0.7135

Framework versions

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