Mohammed Saif AbedandClaude
2fc00efa49
Signed-off-by: mo saif <moisf56@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>
26 lines
927 B
Markdown
26 lines
927 B
Markdown
# MammothModa2-Preview
|
|
|
|
> Text-to-image (T2I) generation has moved to the shared offline image example.
|
|
> See the recipe: [`recipes/MammothModa2/MammothModa2-Preview.md`](../../../recipes/MammothModa2/MammothModa2-Preview.md).
|
|
>
|
|
> This directory retains the image-understanding (image → text) example below,
|
|
> which is outside the scope of the diffusion `model_extras` example migration
|
|
> (#4548 / RFC #4539: AR / Omni understanding examples keep dedicated scripts).
|
|
|
|
## Run example (MammothModa2-Preview)
|
|
|
|
Download model
|
|
```bash
|
|
hf download bytedance-research/MammothModa2-Preview --local-dir ./MammothModa2-Preview
|
|
```
|
|
|
|
### Image Summary
|
|
|
|
```bash
|
|
python examples/offline_inference/mammothmodal2_preview/run_mammothmoda2_image_summarize.py \
|
|
--model ./MammothModa2-Preview \
|
|
--stage-config ./vllm_omni/model_executor/stage_configs/mammoth_moda2_ar.yaml \
|
|
--question "Summarize this image." \
|
|
--image ./image.png
|
|
```
|