Skip to content

add script for training gpt3 on XPU machine using flagcx as comm backend #11014

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: develop
Choose a base branch
from
Open

add script for training gpt3 on XPU machine using flagcx as comm backend #11014

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 40 additions & 0 deletions tests/test_tipc/dygraph/hybrid_parallelism/gpt3/test_flagcx_xpu/run.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
# this is the script for training gpt3 on XPU machines using flagcx as communication backend
export root_path=/workspace
#export PYTHONPATH=$root_path/PaddleNLP:$PYTHONPATH
export PADDLE_DISTRI_BACKEND=flagcx

# log
export GLOG_v=0
export FLAGCX_DEBUG=INFO
export FLAGCX_DEBUG_SUBSYS=INIT
export XPU_FORCE_SHARED_DEVICE_CONTEXT=1
# export FLAGS_log_memory_stats=0
# export GLOG_minloglevel=3
# export TRANSLATOR_CODE_LEVEL=100

current_date=$(date +"%m%d")
task_name="gpt13b_dynamic_hand_nosp_ly4_debug_$current_date"
# task_name="gpt13b_dynamic_baseline_ly20_nosp_nofusedrop$current_date"
# task_name="gpt13b_dynamic_baseline_$current_date"
log_dir="log_$current_date/${task_name}_1"
output_dir="output_$current_date/${task_name}_1"

rm -rf ${log_dir}
rm -rf ${output_dir}


python -u -m paddle.distributed.launch \
--xpus "0,1,2,3,4,5,6,7" \
--log_dir ${log_dir} \
run_pretrain.py \
${root_path}/PaddleNLP/tests/test_tipc/dygraph/hybrid_parallelism/gpt3/auto_config_gpt3_13b/pretrain-gpt3_13b-config.json

echo "---- $task_name performance:"
echo "throughput(tokens/s/card):"
cat ${log_dir}/workerlog.0 | grep "interval_tokens_per_second_per_device:" | awk -F ',' '{print $11}' | awk -F ' ' '{print $2}' | awk 'NR > 10 {print $1}' |sort -n | awk '{values[NR] = $1} END {for (i = 3; i <= NR-2; i++) sum += values[i]; print sum / (NR-4)}'

echo "max_memory_allocated(GB):"
cat ${log_dir}/workerlog.0 | grep "interval_tokens_per_second_per_device:" | awk -F ',' '{print $7}' | tail -n 1

echo "max_memory_reserved(GB):"
cat ${log_dir}/workerlog.0 | grep "interval_tokens_per_second_per_device:" | awk -F ',' '{print $8}' | tail -n 1
Loading