training.md

Fine-tuning FastChat-T5

You can use the following command to train FastChat-T5 with 4 x A100 (40GB).

torchrun --nproc_per_node=4 --master_port=9778 fastchat/train/train_flant5.py \
    --model_name_or_path google/flan-t5-xl \
    --data_path ./data/dummy_conversation.json \
    --bf16 True \
    --output_dir ./checkpoints_flant5_3b \
    --num_train_epochs 3 \
    --per_device_train_batch_size 1 \
    --per_device_eval_batch_size 1 \
    --gradient_accumulation_steps 4 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 300 \
    --save_total_limit 1 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --warmup_ratio 0.03 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --fsdp "full_shard auto_wrap" \
    --fsdp_transformer_layer_cls_to_wrap T5Block \
    --tf32 True \
    --model_max_length 2048 \
    --preprocessed_path ./preprocessed_data/processed.json \
    --gradient_checkpointing True 

After training, please use our post-processing function to update the saved model weight. Additional discussions can be found here.

Fine-tuning using (Q)LoRA

You can use the following command to train Vicuna-7B using QLoRA using ZeRO2. Note that ZeRO3 is not currently supported with QLoRA but ZeRO3 does support LoRA, which has a reference configuraiton under playground/deepspeed_config_s3.json. To use QLoRA, you must have bitsandbytes>=0.39.0 and transformers>=4.30.0 installed.

deepspeed fastchat/train/train_lora.py \
    --model_name_or_path ~/model_weights/llama-7b \
    --lora_r 8 \
    --lora_alpha 16 \
    --lora_dropout 0.05 \
    --data_path ./data/dummy_conversation.json \
    --bf16 True \
    --output_dir ./checkpoints \
    --num_train_epochs 3 \
    --per_device_train_batch_size 1 \
    --per_device_eval_batch_size 1 \
    --gradient_accumulation_steps 1 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 1200 \
    --save_total_limit 100 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --warmup_ratio 0.03 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --tf32 True \
    --model_max_length 2048 \
    --q_lora True \
    --deepspeed playground/deepspeed_config_s2.json \

For T5-XL or XXL

deepspeed fastchat/train/train_lora_t5.py \
        --model_name_or_path google/flan-t5-xl    \
        --data_path ./data/dummy_conversation.json \
        --bf16 True \
        --output_dir ./checkpoints_flant5_3b \
        --num_train_epochs 3 \
        --per_device_train_batch_size 1 \
        --per_device_eval_batch_size 1  \
        --gradient_accumulation_steps 4  \
        --evaluation_strategy "no"  \
        --save_strategy "steps"  \
        --save_steps 300 \
        --save_total_limit 1 \
        --learning_rate 2e-5 \
        --weight_decay 0.     \
        --warmup_ratio 0.03    \
        --lr_scheduler_type "cosine"   \
        --logging_steps 1 \
        --model_max_length 2048    \
        --preprocessed_path ./preprocessed_data/processed.json \
        --gradient_checkpointing True \
        --q_lora True     \
        --deepspeed playground/deepspeed_config_s2.json
        

Fine-tuning Vicuna-7B with Local NPUs

You can use the following command to train Vicuna-7B with 8 x 910B (60GB). Use --nproc_per_node to specify the number of NPUs.

torchrun --nproc_per_node=8 --master_port=20001 fastchat/train/train.py \
    --model_name_or_path ~/vicuna-7b-v1.5-16k  \
    --data_path data/dummy_conversation.json \
    --fp16 True \
    --output_dir output_vicuna \
    --num_train_epochs 3 \
    --per_device_train_batch_size 8 \
    --per_device_eval_batch_size 1 \
    --gradient_accumulation_steps 1 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 1200 \
    --save_total_limit 10 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --warmup_ratio 0.03 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --fsdp "full_shard auto_wrap" \
    --fsdp_transformer_layer_cls_to_wrap 'LlamaDecoderLayer' \
    --model_max_length 2048 \
    --gradient_checkpointing True \
    --lazy_preprocess True