Skip to content

Commit

Permalink
file left out from previous commit + conf typo fix
Browse files Browse the repository at this point in the history
  • Loading branch information
rkarhila-amd committed Nov 15, 2024
1 parent 37b35f7 commit 87b987f
Show file tree
Hide file tree
Showing 2 changed files with 13 additions and 1 deletion.
12 changes: 12 additions & 0 deletions benchmarks/llm/recipes/full_finetune_distributed.py
Original file line number Diff line number Diff line change
@@ -1,3 +1,15 @@
#!/usr/bin/env python3

# As of November 2024, the development of torchrun is very rapid.
# This is the recipe based on torchrun recipe git commit e137afe (post release 0.3.1)
# https://github.com/pytorch/torchtune/blob/7bfb3336446f0d874ab5d4595249839b735b7076/recipes/lora_finetune_distributed.py

# Torchtune 0.2.1 recipe with device instrumenation (c) Mila
# https://github.com/mila-iqia/milabench/blob/a60a3aae21e87e46bcce403620a3f56c12878554/benchmarks/llm/recipes/full_finetune_distributed.py

# The instrumentation edits (c) AMD


# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
Expand Down
2 changes: 1 addition & 1 deletion config/base.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -652,7 +652,7 @@ llm-full-mp-gpus:

argv:
#"{milabench_code}/recipes/full_finetune_distributed.py": true
tuneworkaroundrecipes.full_finetune_distributed: true
tuneworkaroundrecipes/full_finetune_distributed.py: true
--config: "{milabench_code}/configs/llama3_70B_full.yaml"
epochs=1: true
output_dir={milabench_extra}/output: true
Expand Down

0 comments on commit 87b987f

Please sign in to comment.