lmind_hotpot_train8000_eval7405_v1_recite_qa_Qwen_Qwen1.5-4B_5e-4_lora2
This model is a fine-tuned version of Qwen/Qwen1.5-4B on the tyzhu/lmind_hotpot_train8000_eval7405_v1_recite_qa dataset. It achieves the following results on the evaluation set:
- Loss: 0.5757
- Accuracy: 0.7612
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: 0.0005
- 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 |
---|---|---|---|---|
1.5463 | 0.9998 | 1089 | 1.3539 | 0.6872 |
1.3199 | 1.9995 | 2178 | 1.1632 | 0.7022 |
1.1039 | 2.9993 | 3267 | 1.0347 | 0.7134 |
0.9356 | 4.0 | 4357 | 0.9234 | 0.7237 |
0.8312 | 4.9998 | 5446 | 0.8529 | 0.7307 |
0.7565 | 5.9995 | 6535 | 0.7860 | 0.7372 |
0.6985 | 6.9993 | 7624 | 0.7415 | 0.7415 |
0.6623 | 8.0 | 8714 | 0.7111 | 0.7457 |
0.6281 | 8.9998 | 9803 | 0.6775 | 0.7481 |
0.5885 | 9.9995 | 10892 | 0.6689 | 0.7496 |
0.5721 | 10.9993 | 11981 | 0.6364 | 0.7530 |
0.5504 | 12.0 | 13071 | 0.6319 | 0.7541 |
0.5406 | 12.9998 | 14160 | 0.6185 | 0.7549 |
0.536 | 13.9995 | 15249 | 0.6158 | 0.7565 |
0.5205 | 14.9993 | 16338 | 0.5976 | 0.7578 |
0.5175 | 16.0 | 17428 | 0.5922 | 0.7590 |
0.5068 | 16.9998 | 18517 | 0.5823 | 0.7593 |
0.5023 | 17.9995 | 19606 | 0.5754 | 0.7607 |
0.4848 | 18.9993 | 20695 | 0.5781 | 0.7608 |
0.4767 | 19.9954 | 21780 | 0.5757 | 0.7612 |
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
- 2
Model tree for tyzhu/lmind_hotpot_train8000_eval7405_v1_recite_qa_Qwen_Qwen1.5-4B_5e-4_lora2
Base model
Qwen/Qwen1.5-4B
Adapter
this model
Dataset used to train tyzhu/lmind_hotpot_train8000_eval7405_v1_recite_qa_Qwen_Qwen1.5-4B_5e-4_lora2
Evaluation results
- Accuracy on tyzhu/lmind_hotpot_train8000_eval7405_v1_recite_qaself-reported0.761