Open
Conversation
sarapapi
reviewed
Mar 17, 2026
| @@ -0,0 +1,119 @@ | |||
| # Copyright 2025 FBK, KIT | |||
Collaborator
There was a problem hiding this comment.
Suggested change
| # Copyright 2025 FBK, KIT | |
| # Copyright 2026 FBK, KIT |
| torch_dtype="auto", | ||
| device_map="auto", | ||
| attn_implementation="flash_attention_2", | ||
| ) |
Collaborator
There was a problem hiding this comment.
In the model card, they suggest adding model.disable_talker() when the model is not used for speech generation (as in our case). Not sure if it has a real impact on performance tbh, but worth to try
| "content": [ | ||
| { | ||
| "type": "text", | ||
| "text": "You are Qwen, a virtual human developed by the Qwen Team, Alibaba Group, " |
Collaborator
There was a problem hiding this comment.
Is it the standard one, right?
| ) | ||
| inputs = inputs.to(model.device).to(model.dtype) | ||
|
|
||
| # Inference: Generation of the output text |
Collaborator
There was a problem hiding this comment.
Suggested change
| # Inference: Generation of the output text |
Collaborator
There was a problem hiding this comment.
Not really necessary here, we only do inference in this repo
Collaborator
|
Have you already tried to run it, right? |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds Qwen3-Omni-30B-A3B-Instruct as a MLLM baseline under the name
qwen3_omni.Changes
baselines/models/mllm/qwen3_omni.pybaselines/main.pyExample usage
python main.py \ --model qwen3_omni \ --lang en \ --track long \ --modality audio \ --prompt fixed \ --in_data_folder $PATH2DATA \ --out_folder ./baselineNotes
flash_attention_2and sufficient VRAM for a 30B MoE model