lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_3e-4_lora2

This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_hotpot_train8000_eval7405_v1_qa dataset. It achieves the following results on the evaluation set:

  • Loss: 3.6823
  • Accuracy: 0.4917

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: 0.0003
  • 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
2.248 1.0 250 2.3110 0.5173
1.9103 2.0 500 2.3740 0.5157
1.4896 3.0 750 2.5266 0.5112
1.109 4.0 1000 2.7830 0.5037
0.7757 5.0 1250 3.0311 0.4987
0.5994 6.0 1500 3.2256 0.4979
0.4921 7.0 1750 3.3517 0.4958
0.4575 8.0 2000 3.4321 0.4946
0.4233 9.0 2250 3.5151 0.4961
0.4178 10.0 2500 3.5280 0.4950
0.3987 11.0 2750 3.5547 0.4951
0.4033 12.0 3000 3.5601 0.4954
0.3932 13.0 3250 3.5859 0.4932
0.4012 14.0 3500 3.5944 0.4927
0.3895 15.0 3750 3.6038 0.4939
0.396 16.0 4000 3.6504 0.4932
0.3847 17.0 4250 3.6602 0.4912
0.3942 18.0 4500 3.6515 0.4914
0.3809 19.0 4750 3.7304 0.4923
0.3805 20.0 5000 3.6823 0.4917

Framework versions

  • PEFT 0.5.0
  • Transformers 4.40.2
  • Pytorch 2.3.0
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
3
Inference API
Unable to determine this model’s pipeline type. Check the docs .

Model tree for tyzhu/lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_3e-4_lora2

Base model

Qwen/Qwen1.5-4B
Adapter
(272)
this model

Dataset used to train tyzhu/lmind_hotpot_train8000_eval7405_v1_qa_Qwen_Qwen1.5-4B_3e-4_lora2

Evaluation results

  • Accuracy on tyzhu/lmind_hotpot_train8000_eval7405_v1_qa
    self-reported
    0.492