CMC-Deepseek-14B-legal-v5

This model is a fine-tuned version of /home/htdat/huggingface/deepseek-ai/DeepSeek-R1-Distill-Qwen-14B on the thanhdath/legal-summarize, the CATI-AI/legal_conflict_distilation_v3, the CATI-AI/vietnamese_legal_closed_qa_cot, the CATI-AI/summary_data_hf_dataset_markdown, the CATI-AI/closed-qa-document-pairs-markdown and the CATI-AI/legal_summarization_markdown_v2 datasets. It achieves the following results on the evaluation set:

• Loss: nan

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: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 7
• gradient_accumulation_steps: 16
• total_train_batch_size: 112
• total_eval_batch_size: 7
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 2

Training results

Framework versions

• Transformers 4.48.1
• Pytorch 2.2.2+cu121
• Datasets 4.2.0
• Tokenizers 0.21.4