The data generation process utilizes the Alpaca Self-Instruct pipeline, but with the OpenAI chat model (e.g., gpt-3.5-turbo
). Please note tthat his pipeline does not employ a batch system since the chat model does not support prompt batching.
python -m generate_data generate_instruction_following_data \
--output_dir="./" \
--num_instructions_to_generate=100 \
--num_instructions_to_generate_per_request=2 \
--model_name="gpt-3.5-turbo-16k" \
--similarity_threshold=0.6
The fine-tuning process utilizes the WizardLM-13B-V1.2 model.
python finetune.py \
--hf_token hf_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx \
--base_model_id codellama/CodeLlama-7b-hf \
--dataset_id alimtegar/indowebgen-dataset \
--output_dir "./indowebgen-7b-4k-lora" \
--output_model_id alimtegar/indowebgen-7b-4k-lora \
--commit_message "Finetune for 3 epochs" \
--vastai_api_key xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx \
--vastai_instance_id 1234567
--stop_vastai_instance 0