Skip to content

delete generate_rank_mapping when export multi cards model #9552

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
ZHUI merged 1 commit into from
Dec 3, 2024
Merged

delete generate_rank_mapping when export multi cards model #9552

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
delete generate_rank_mapping when export multi cards model
  • Loading branch information
@yuanlehome
yuanlehome committed Dec 3, 2024
commit d5728e062cf5d6da0006b811894e940399939c0b
1 change: 0 additions & 1 deletion llm/predict/export_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,6 @@ def main():
predictor.model.generation_config.save_pretrained(export_args.output_path)

predictor.tokenizer.save_pretrained(export_args.output_path)
llm_utils.generate_rank_mapping(os.path.join(export_args.output_path, "rank_mapping.csv"))

if tensor_parallel_degree > 1:
export_args.output_path = os.path.join(export_args.output_path, f"rank_{tensor_parallel_rank}")
Expand Down
23 changes: 0 additions & 23 deletions paddlenlp/trl/llm_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -267,29 +267,6 @@ def get_infer_model_path(input_dir, model_prefix):
return os.path.join(input_dir, model_prefix)


def generate_rank_mapping(output_filename):
ring_id = -1
try:
hcg = fleet.get_hybrid_communicate_group()
model_parallel_group = hcg.get_model_parallel_group()
ring_id = model_parallel_group.id
except Exception:
pass

if ring_id == -1:
return

world_size = dist.get_world_size()
with open(output_filename, "w") as f:
f.write("[ring_id -> ranks]\n")
f.write(",".join(map(str, [0] + list(range(world_size)))) + "\n")
f.write(",".join(map(str, [ring_id] + list(range(world_size)))) + "\n")

f.write("[rank -> ring_ids]\n")
for i in range(world_size):
f.write("{},0,{}\n".format(i, ring_id))


def deserialize_from_file(fp):
x_type = fp.read(1)
x_type_out = struct.unpack("c", x_type)[0]
Expand Down