chenhaodev/internlm2-20b-llama-onc-v1

model-update

This model is a fine-tuned version of chargoddard/internlm2-20b-llama on the oncc_medqa_instruct dataset.

Training procedure

CUDA_VISIBLE_DEVICES=0 python src/train_bash.py --stage sft --do_train True --model_name_or_path /workspace/model --finetuning_type lora --quantization_bit 4 --flash_attn True --dataset_dir data --cutoff_len 1024 --learning_rate 0.0005 --num_train_epochs 1.0 --max_samples 10000 --lr_scheduler_type cosine --max_grad_norm 1.0 --logging_steps 10 --save_steps 100 --warmup_steps 20 --neftune_noise_alpha 0.5 --lora_rank 8 --lora_dropout 0.2 --output_dir /workspace/model-update --per_device_train_batch_size 4 --gradient_accumulation_steps 4 --lora_target q_proj,v_proj --template llama2 --dataset oncc_medqa_instruct

Note: fix the bug in the tokenizer_config.json. i.e. "internlm/internlm2-20b--tokenization_internlm2.InternLM2Tokenizer"

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 20
• num_epochs: 1.0

Training results

Framework versions

• PEFT 0.8.2
• Transformers 4.37.2
• Pytorch 2.0.1+cu118
• Datasets 2.17.0
• Tokenizers 0.15.2