From bd861a14ad1c69c53348b87827d3ae32c8a36e5e Mon Sep 17 00:00:00 2001 From: Tong Li Date: Fri, 12 Jul 2024 06:30:26 +0000 Subject: [PATCH] fix issue --- .../ColossalChat/examples/training_scripts/train_dpo.sh | 2 +- .../ColossalChat/examples/training_scripts/train_orpo.sh | 4 ++-- .../ColossalChat/examples/training_scripts/train_ppo.sh | 2 +- .../ColossalChat/examples/training_scripts/train_rm.sh | 2 +- .../ColossalChat/examples/training_scripts/train_sft.sh | 2 +- 5 files changed, 6 insertions(+), 6 deletions(-) diff --git a/applications/ColossalChat/examples/training_scripts/train_dpo.sh b/applications/ColossalChat/examples/training_scripts/train_dpo.sh index f7bb456584c2..48c5f7181288 100755 --- a/applications/ColossalChat/examples/training_scripts/train_dpo.sh +++ b/applications/ColossalChat/examples/training_scripts/train_dpo.sh @@ -38,7 +38,7 @@ declare -a dataset=( TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S) FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}" SAVE_DIR="${PARENT_SAVE_DIR}${FULL_PROJECT_NAME}" -CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json" +CONFIG_FILE="${PARENT_CONFIG_FILE}${FULL_PROJECT_NAME}.json" colossalai run --nproc_per_node 4 --hostfile hostfile --master_port 31313 train_dpo.py \ --pretrain $PRETRAINED_MODEL_PATH \ diff --git a/applications/ColossalChat/examples/training_scripts/train_orpo.sh b/applications/ColossalChat/examples/training_scripts/train_orpo.sh index ca80a14c1f7d..1253fa71988c 100755 --- a/applications/ColossalChat/examples/training_scripts/train_orpo.sh +++ b/applications/ColossalChat/examples/training_scripts/train_orpo.sh @@ -13,7 +13,7 @@ set_n_least_used_CUDA_VISIBLE_DEVICES() { echo "Now CUDA_VISIBLE_DEVICES is set to:" echo "CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES" } -set_n_least_used_CUDA_VISIBLE_DEVICES 8 +set_n_least_used_CUDA_VISIBLE_DEVICES 2 PROJECT_NAME="dpo" PARENT_SAVE_DIR="" # Path to a folder to save checkpoints @@ -38,7 +38,7 @@ declare -a dataset=( TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S) FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}" SAVE_DIR="${PARENT_SAVE_DIR}${FULL_PROJECT_NAME}" -CONFIG_FILE="${PARENT_CONFIG_FILE}-${FULL_PROJECT_NAME}.json" +CONFIG_FILE="${PARENT_CONFIG_FILE}${FULL_PROJECT_NAME}.json" colossalai run --nproc_per_node 8 --hostfile hostfile --master_port 31313 train_orpo.py \ --pretrain $PRETRAINED_MODEL_PATH \ diff --git a/applications/ColossalChat/examples/training_scripts/train_ppo.sh b/applications/ColossalChat/examples/training_scripts/train_ppo.sh index 91633978e6ff..559f630a8835 100755 --- a/applications/ColossalChat/examples/training_scripts/train_ppo.sh +++ b/applications/ColossalChat/examples/training_scripts/train_ppo.sh @@ -54,7 +54,7 @@ declare -a ptx_dataset=( TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S) FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}" 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" colossalai run --nproc_per_node 8 --hostfile hostfile --master_port 31312 train_ppo.py \ --pretrain $PRETRAINED_MODEL_PATH \ diff --git a/applications/ColossalChat/examples/training_scripts/train_rm.sh b/applications/ColossalChat/examples/training_scripts/train_rm.sh index e06d9092fe4c..571c650a489c 100755 --- a/applications/ColossalChat/examples/training_scripts/train_rm.sh +++ b/applications/ColossalChat/examples/training_scripts/train_rm.sh @@ -38,7 +38,7 @@ declare -a dataset=( TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S) FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}" 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" colossalai run --nproc_per_node 8 --hostfile hostfile --master_port 31312 train_rm.py \ --pretrain $PRETRAINED_MODEL_PATH \ diff --git a/applications/ColossalChat/examples/training_scripts/train_sft.sh b/applications/ColossalChat/examples/training_scripts/train_sft.sh index 18df0929327e..678bfc809819 100755 --- a/applications/ColossalChat/examples/training_scripts/train_sft.sh +++ b/applications/ColossalChat/examples/training_scripts/train_sft.sh @@ -36,7 +36,7 @@ declare -a dataset=( TIMESTAMP=$(date +%Y-%m-%d-%H-%M-%S) FULL_PROJECT_NAME="${PROJECT_NAME}-${TIMESTAMP}" 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)