mirror of https://github.com/hpcaitech/ColossalAI
fix style
parent
d888c3787c
commit
f6ef5c3609
|
@ -17,16 +17,14 @@ set_n_least_used_CUDA_VISIBLE_DEVICES 4
|
||||||
|
|
||||||
PROJECT_NAME="dpo"
|
PROJECT_NAME="dpo"
|
||||||
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
||||||
PRETRAINED_MODEL_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local model path
|
PRETRAINED_MODEL_PATH="" # huggingface or local model path
|
||||||
PRETRAINED_TOKENIZER_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local tokenizer path
|
PRETRAINED_TOKENIZER_PATH="" # huggingface or local tokenizer path
|
||||||
|
|
||||||
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
||||||
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
||||||
SAVE_DIR="${PARENT_SAVE_DIR}${FULL_PROJECT_NAME}"
|
SAVE_DIR="${PARENT_SAVE_DIR}${FULL_PROJECT_NAME}"
|
||||||
CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json"
|
CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json"
|
||||||
|
|
||||||
echo $(which colossalai)
|
|
||||||
echo $(which python)
|
|
||||||
colossalai run --nproc_per_node 4 --master_port 31313 benchmark_dpo.py \
|
colossalai run --nproc_per_node 4 --master_port 31313 benchmark_dpo.py \
|
||||||
--pretrain $PRETRAINED_MODEL_PATH \
|
--pretrain $PRETRAINED_MODEL_PATH \
|
||||||
--tokenizer_dir $PRETRAINED_TOKENIZER_PATH \
|
--tokenizer_dir $PRETRAINED_TOKENIZER_PATH \
|
||||||
|
|
|
@ -17,8 +17,8 @@ set_n_least_used_CUDA_VISIBLE_DEVICES 2
|
||||||
|
|
||||||
PROJECT_NAME="dpo"
|
PROJECT_NAME="dpo"
|
||||||
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
||||||
PRETRAINED_MODEL_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local model path
|
PRETRAINED_MODEL_PATH="" # huggingface or local model path
|
||||||
PRETRAINED_TOKENIZER_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local tokenizer path
|
PRETRAINED_TOKENIZER_PATH="" # huggingface or local tokenizer path
|
||||||
|
|
||||||
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
||||||
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
||||||
|
|
|
@ -17,15 +17,13 @@ set_n_least_used_CUDA_VISIBLE_DEVICES 4
|
||||||
# export CUDA_VISIBLE_DEVICES=3,4
|
# export CUDA_VISIBLE_DEVICES=3,4
|
||||||
PROJECT_NAME="sft"
|
PROJECT_NAME="sft"
|
||||||
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
PARENT_CONFIG_FILE="./benchmark_config" # Path to a folder to save training config logs
|
||||||
PRETRAINED_MODEL_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local model path
|
PRETRAINED_MODEL_PATH="" # huggingface or local model path
|
||||||
PRETRAINED_TOKENIZER_PATH="/root/commonData/Llama-2-7b-hf" # huggingface or local tokenizer path
|
PRETRAINED_TOKENIZER_PATH="" # huggingface or local tokenizer path
|
||||||
|
|
||||||
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S)
|
||||||
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}"
|
||||||
CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json"
|
CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json"
|
||||||
|
|
||||||
echo $(which colossalai)
|
|
||||||
echo $(which python)
|
|
||||||
# the real batch size for gradient descent is number_of_node_in_hostfile * nproc_per_node * train_batch_size
|
# the real batch size for gradient descent is number_of_node_in_hostfile * nproc_per_node * train_batch_size
|
||||||
colossalai run --nproc_per_node 4 --master_port 31312 benchmark_sft.py \
|
colossalai run --nproc_per_node 4 --master_port 31312 benchmark_sft.py \
|
||||||
--pretrain $PRETRAINED_MODEL_PATH \
|
--pretrain $PRETRAINED_MODEL_PATH \
|
||||||
|
|
Loading…
Reference in New Issue