Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Shark FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model #364

Open
kumardeepakamd opened this issue Jan 23, 2024 · 0 comments
Open
Assignees

Comments

@kumardeepakamd
Copy link

kumardeepakamd commented Jan 23, 2024

ONNX to LinAlg Lowering

PyTorch Issues:
[#2832] Unable to export int8/4 ONNX model for OPT/LLAMA2 PyTorch 119621

Issues with lowering to LinAlg:
[#2838] bfloat16 MLP model IREE issue (@PhaneeshB )

Mismatch
[#430] int8 Resnet50 mismatch (@rsuderman ) -- Rob investigated it and found that actual classification is fine
#431 bfloat16 conv2d/resnet50 mismatch (@PhaneeshB )

PyTorch to LinAlg Lowering

Fx Importer
[#2843] Fx Importer does not support bfloat16 (@dan-garvey )

@kumardeepakamd kumardeepakamd self-assigned this Jan 23, 2024
@kumardeepakamd kumardeepakamd changed the title VAI-ML-FE : Support opt-1.3B, laama2-7b, ResNet50 ONNX and PyTorch Models VAI-ML-FE : Support opt-1.3B, laama2-7b, ResNet50 ONNX to LinAlg lowering Jan 31, 2024
@kumardeepakamd kumardeepakamd changed the title VAI-ML-FE : Support opt-1.3B, laama2-7b, ResNet50 ONNX to LinAlg lowering VAI-ML-FE : Support opt-1.3B, laama2-7b, ResNet50 ONNX or PyTorch to LinAlg lowering Jan 31, 2024
@kumardeepakamd kumardeepakamd changed the title VAI-ML-FE : Support opt-1.3B, laama2-7b, ResNet50 ONNX or PyTorch to LinAlg lowering VAI-ML-FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model in FE and IREE Feb 9, 2024
@kumardeepakamd kumardeepakamd changed the title VAI-ML-FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model in FE and IREE Shark FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model in FE and IREE Feb 12, 2024
@kumardeepakamd kumardeepakamd changed the title Shark FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model in FE and IREE Shark FE : Support bfloat16/int8 opt/laama2-7b Fx and ONNX model Feb 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
Status: In Progress
Development

No branches or pull requests

1 participant