Support ONNX conversion and pipeline for SD3 #8984
Conversation
Signed-off-by: Mengni Wang <mengni.wang@intel.com>
Signed-off-by: Mengni Wang <mengni.wang@intel.com>
| scheduler=pipeline.scheduler, | ||
| ) | ||
|
|
||
| onnx_pipeline.save_pretrained("sd3") |
There was a problem hiding this comment.
Since I have no access to upload ONNX SD3 model to hf-internal-testing, I export the ONNX model in UT. If anyone could help upload it, the code can be simplified.
|
@a-r-r-o-w Hi, could you please help to take a look? |
|
I think this should go to Optimum. Cc: @echarlaix |
|
I found SD ONNX conversion and ONNX pipeline are in diffusers so I added ONNX SD3 here. If you all think Optimum is more suitable, please guide me how should I organize it since there seems no ONNX SD related conversion script and pipeline in https://github.com/huggingface/optimum 😄 |
|
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Please note that issues that do not follow the contributing guidelines are likely to be ignored. |
| "input_ids": {0: "batch", 1: "sequence"}, | ||
| }, | ||
| opset=opset, | ||
| ) |
There was a problem hiding this comment.
Can you also collate the external tensor files into one file for this model? Like what you are doing with the transformer.
|
Hello, we have added SD3 and Flux support in huggingface/optimum#2073. |
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
|
closing this now because we don't need this PR anymore because it is already in Optimum let me know if we still need it and I can re-open it! |
What does this PR do?
Support ONNX conversion and pipeline for SD3