This repository was archived by the owner on Sep 10, 2025. It is now read-only.
Multiturn mm single image #1270
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…quence length so their shapes are both set to the max_seq_len
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchchat/1270
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit d494aa0 with merge base d8c0aaf ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Jack-Khuu
suggested changes
Oct 4, 2024
| if len(prompt.shape) > 1: | ||
| prompt = prompt.squeeze(0) | ||
| T = prompt.size(0) | ||
| max_new_tokens = min(max_new_tokens, max_seq_length - start_pos - T) |
Contributor
There was a problem hiding this comment.
Is this line necessary to remove?
Contributor
There was a problem hiding this comment.
This seems like it would be problem with long prompt_lengths
| ] | ||
| image_found = False | ||
| messages = [] | ||
| for message in prompt: |
Contributor
There was a problem hiding this comment.
This would be torchchat.py generate Llama3.2-11B right?
Since it sends prompt: str and uses the image_prompt field
You might need to "create" a container prompt with those 2
Contributor
There was a problem hiding this comment.
Or chat calls a curried version of this function that creates the format before calling this function
Comment on lines
316
to
+334
| if not isinstance(self.model, FlamingoModel): | ||
| prompt = [ | ||
| {"role": message["role"], "content": message["content"]} | ||
| for message in completion_request.messages | ||
| for message in messages | ||
| ] | ||
| return self._gen_model_input( | ||
| prompt=prompt, max_new_tokens=completion_request.max_tokens | ||
| ) | ||
|
|
||
| # Llama 3.2 11B | ||
| prompt = None | ||
| images = None | ||
|
|
||
| for message in messages: | ||
| torchtune_contents = [] | ||
| if isinstance(message["content"], list): | ||
| for content_dict in message["content"]: | ||
| if content_dict["type"] == "text": | ||
| assert ( | ||
| prompt is None | ||
| ), "At most one text prompt is supported for each request" | ||
| prompt = content_dict["text"] | ||
| elif content_dict["type"] == "image_url": | ||
| assert ( | ||
| images is None | ||
| ), "At most one image is supported at the moment" | ||
|
|
||
| base64_decoded = base64.b64decode( | ||
| content_dict["image_url"].split(";base64,")[1] | ||
| ) | ||
| images = [Image.open(BytesIO(base64_decoded))] | ||
|
|
||
| assert prompt is not None, "Text prompt must be specified in the request" | ||
|
|
||
| return self._gen_model_input(prompt, images, completion_request.max_tokens) | ||
|
|
||
| prompt = [ | ||
| {"role": message["role"], "content": message["content"]} | ||
| for message in messages | ||
| ] | ||
|
|
||
| return self._gen_model_input( | ||
| prompt=prompt, max_new_tokens=completion_request.max_tokens | ||
| ) |
vmpuri
force-pushed
the
multiturn-mm-single-image
branch
from
f5b512e to
be0632b
Compare
vmpuri
force-pushed
the
multiturn-mm-single-image
branch
from
be0632b to
d494aa0
Compare
Jack-Khuu
approved these changes
Oct 5, 2024
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Multi-turn conversations were not working within the browser for LLaMA 3.2 Vision. This fixes this for the case of:
Tests
CLI Tests
Tested with server + browser:
python3 torchchat.py server llama3.2-11bstreamlit run torchchat/usages/browser.pySingle Image
Text Only