Skip to content

fix training issue on NVIDIA GPUs #473

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

Merged
Xiaoming-AMD merged 1 commit into from
Jan 6, 2026
Merged

fix training issue on NVIDIA GPUs #473

Show file tree
Hide file tree
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
Empty file added examples/__init__.py
View file
Open in desktop
Empty file.
2 changes: 1 addition & 1 deletion examples/run_pretrain.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -506,7 +506,7 @@ fi
setup_pythonpath() {
local site_packages
site_packages=$(python -c "import sysconfig; print(sysconfig.get_paths()['purelib'])")
export PYTHONPATH="${site_packages}:${PRIMUS_PATH}:$:${PYTHONPATH}"
export PYTHONPATH="${PRIMUS_PATH}:${site_packages}:${PYTHONPATH}"
}

setup_pythonpath
Expand Down
Empty file added examples/scripts/__init__.py
View file
Open in desktop
Empty file.
12 changes: 9 additions & 3 deletions primus/modules/trainer/megatron/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1957,7 +1957,9 @@ def training_log(
# Note(wenx): If we want to collect rocm-smi memory information for the first two iterations,
# place the collection before the timer to minimize its impact on latency measurements for iterations ≥ 3.
if args.log_throughput:
if args.use_rocm_mem_info or iteration in args.use_rocm_mem_info_iters:
if args.use_rocm_mem_info or (
args.use_rocm_mem_info_iters is not None and iteration in args.use_rocm_mem_info_iters
):
rocm_total_mem, rocm_used_mem, rocm_free_mem = get_rocm_smi_mem_info(
self.module_local_rank
)
Expand Down Expand Up @@ -2015,7 +2017,9 @@ def training_log(
log_string += f"{hip_free_mem/1024/1024/1024:.2f}GiB/"
log_string += f"{hip_total_mem/1024/1024/1024:.2f}GiB/{hip_mem_usage*100:.2f}% |"

if args.use_rocm_mem_info or iteration in args.use_rocm_mem_info_iters:
if args.use_rocm_mem_info or (
args.use_rocm_mem_info_iters is not None and iteration in args.use_rocm_mem_info_iters
):
rocm_mem_usage = rocm_used_mem / rocm_total_mem

# get the max rocm_mem_usage
Expand Down Expand Up @@ -2052,7 +2056,9 @@ def training_log(
f"{statistics.mean(self.recent_token_throughputs):.1f} |"
)
if args.log_timers_to_tensorboard:
if args.use_rocm_mem_info or iteration in args.use_rocm_mem_info_iters:
if args.use_rocm_mem_info or (
args.use_rocm_mem_info_iters is not None and iteration in args.use_rocm_mem_info_iters
):
mem_collector = "rocm"
used_mem, free_mem, total_mem, mem_usage = (
rocm_used_mem,
Expand Down