lmind_nq_train6000_eval6489_v1_recite_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_recite_qa_v3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4649
• Accuracy: 0.7967

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: 20.0

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
1.8276	1.0	529	0.6129	1.6297
1.6971	2.0	1058	0.6290	1.5016
1.4581	3.0	1587	0.6485	1.3573
1.2293	4.0	2116	0.6700	1.1989
0.9705	5.0	2645	0.6936	1.0435
0.7526	6.0	3174	0.7159	0.8995
0.5841	7.0	3703	0.7358	0.7834
0.4653	8.0	4232	0.7514	0.6874
0.3755	9.0	4761	0.7667	0.5926
0.3156	10.0	5290	0.7757	0.5410
0.2652	11.0	5819	0.7829	0.5042
0.2319	12.0	6348	0.4819	0.7874
0.2047	13.0	6877	0.4747	0.7902
0.1889	14.0	7406	0.4667	0.7927
0.1728	15.0	7935	0.4688	0.7930
0.162	16.0	8464	0.4614	0.7947
0.1512	17.0	8993	0.4588	0.7958
0.1478	18.0	9522	0.4614	0.7957
0.1427	19.0	10051	0.4582	0.7970
0.1371	20.0	10580	0.4649	0.7967

Framework versions

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