Fine-tuned Model: Legal-gemma3-27b-it-

📚 Training Configuration

• data_path: QomSSLab/Legal_DS_SFT
• output_dir: gemma327b_lora_chckpnts
• new_model_name: Legal-gemma3-27b-it-
• model_name: /home/aiuser/SmartJudge/PT/Legal-gemma3-27b-pt-lora
• use_4bit: False
• use_lora: True
• max_seq_length: 3000
• batch_size: 1
• gradient_accu: 4
• epochs: 2
• learning_rate: 0.0001
• lora_alpha: 128
• lora_drop: 0.05
• lora_r: 128
• tune_embedding_layer: False
• hf_token: ********
• resume_from_checkpoint: True
• use_8bit_optimizer: True
• push_to_hub: True