-
mi300_all_sym_8_step14_fp32 updates.
Stella Laurenzo
-
Add model experiment switching to qkv symmetric quant.
Stella Laurenzo
Labeled "qkv_sym" in the experiment list.
-
Update unet assets for fused qkv quantization.
Stella Laurenzo
This is based off of code changes: https://github.com/nod-ai/sharktank/pull/106
-
Initial commit of FP16 vae model.
Stella Laurenzo
-
Upload updated export irpa.
Stella Laurenzo
-
Update to 20240702 drop from Brevitas.
Stella Laurenzo
The primary change is to store the quant params with zero points as int8. This has to be coordinated with a sharktank update. -
Update with conv2d bugfix: https://github.com/nod-ai/sharktank/commit/0b67e6aca5185fc517814bd631f821798e1ee26b
Stella Laurenzo
-
Add sharktank exported assets.
Stella Laurenzo
-
Initial add of unet/int8 model.
Stella Laurenzo
-
Update README.md
-
initial commit
