Tags: quic/aimet
Tags
Fix failing fp16 QDQ export tests Signed-off-by: Michael Tuttle <[email protected]> Co-authored-by: Michael Tuttle <[email protected]>
cherry pick adascale changes from develop (#5739) * Changes for experimental feature AdaScale in aimet-onnx Signed-off-by: Abhi Khobare <[email protected]> * Updated release notes and docs links Signed-off-by: Abhi Khobare <[email protected]> --------- Signed-off-by: Abhi Khobare <[email protected]> Co-authored-by: Abhi Khobare <[email protected]>
Add dummy action to make it easy to play with GH workflow (#5684) Signed-off-by: Bhushan Sonawane <[email protected]> Co-authored-by: Bhushan Sonawane <[email protected]>
[cherry-pick] add e2e accuracy tests for aimet-tf (#5640) * add e2e accuracy tests for aimet-tf Signed-off-by: Bhushan Sonawane <[email protected]> --------- Signed-off-by: Bhushan Sonawane <[email protected]> Co-authored-by: Bhushan Sonawane <[email protected]>
Pin deepspeed to stabilize CI pipeline (#5602) Signed-off-by: Michael Tuttle <[email protected]> Signed-off-by: Bhushan Sonawane <[email protected]> Co-authored-by: Michael Tuttle <[email protected]>
Adjust weight scale for int32 bias overflow in W16A16 quantization Automatically adjust weight scale for int32 bias overflow in INT16 quantization Signed-off-by: Hitarth Mehta <[email protected]> Co-authored-by: Hitarth Mehta <[email protected]>
Work-around onnx version converter issue for models with external (#5… …371) weights If onnx version converter fails, upgrade and map weights separately resolves #5367 Signed-off-by: Bhushan Sonawane <[email protected]> Co-authored-by: Bhushan Sonawane <[email protected]>
AdaScale: Maintain LR per model and apply it * Remove hardcoding to the first sample during loss computation * Fix tuples being passed to loss func Signed-off-by: yathindra kota <[email protected]> Co-authored-by: yathindra kota <[email protected]>
PreviousNext