Skip to content
This repository was archived by the owner on Sep 10, 2025. It is now read-only.

Update torchtune pin to Llama 3.2 Release #1205

Merged
merged 3 commits into from
Sep 25, 2024
Merged

Update torchtune pin to Llama 3.2 Release #1205

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0c3dad3
Update torchtune pin to Llama 3.2 Release
Jack-Khuu Sep 25, 2024
fc731f3
Update flamingo tune components to 3_2
Jack-Khuu Sep 25, 2024
0551988
Bumping the lm_eval install to circumvent tune dependency
Jack-Khuu Sep 25, 2024
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
2 changes: 1 addition & 1 deletion install/install_requirements.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,7 +90,7 @@ REQUIREMENTS_TO_INSTALL=(
# Rely on the latest tochtune for flamingo support
(
set -x
$PIP_EXECUTABLE install git+https://github.com/pytorch/torchtune.git@18efc81dda1c537bb7c25058ff059b4623ccff58
$PIP_EXECUTABLE install git+https://github.com/pytorch/torchtune.git@d002d45e3ec700fa770d9dcc61b02c59e2507bf6
)

if [[ -x "$(command -v nvidia-smi)" ]]; then
Expand Down
11 changes: 6 additions & 5 deletions torchchat/usages/openai_api.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,11 +17,10 @@

import torch

from torchtune.models.llama3_2_vision._convert_weights import padded_collate
from torchtune.models.llama3_2_vision._model_builders import llama3_2_vision_transform

from PIL import Image
from torchtune.data import Message

from torchtune.data import Message, padded_collate
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we need to double check about padded_collate func here; in the generation.py we've replaced it with padded_collate_tiled_images_and_mask

from torchtune.models.llama3_2_vision._model_builders import llama3_2_vision_transform

from torchchat.cli.download import is_model_downloaded, load_model_configs
from torchchat.generate import Generator, GeneratorArgs
Expand Down Expand Up @@ -374,7 +373,9 @@ def chunked_completion(self, completion_request: CompletionRequest):
images.append(Image.open(BytesIO(base64_decoded)))
print("images:", len(images), flush=True)
if len(images) > 0:
transform = llama3_2_vision_transform(str(self.tokenizer_args.tokenizer_path))
transform = llama3_2_vision_transform(
str(self.tokenizer_args.tokenizer_path)
)
torchtune_messages = self._openai_messages_to_torchtune(
completion_request.messages
)
Expand Down
Loading