Require strict profitability before advancing quantization above slices - #22807
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22807
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 1 PendingAs of commit fa0087a with merge base 13ec99a ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
|
@DrJessop has exported this pull request. If you are a Meta employee, you can view the originating Diff in D119550687. |
This PR needs a
|
Summary: Require a strict profitability win before advancing quantization above slice or select operations. When the float slice outputs and the quantized input have equal estimated memory cost, keep the slice before quantization to avoid quantizing elements that will be discarded. Add coverage for the half-sized-slice boundary. Differential Revision: D119550687
meta-codesync
Bot
changed the title
146bde0 to
fa0087a
Compare
…es (pytorch#22807) Summary: Require a strict profitability win before advancing quantization above slice or select operations. When the float slice outputs and the quantized input have equal estimated memory cost, keep the slice before quantization to avoid quantizing elements that will be discarded. Add coverage for the half-sized-slice boundary. Differential Revision: D119550687
2 participants
Summary: Require a strict profitability win before advancing quantization above slice or select operations. When the float slice outputs and the quantized input have equal estimated memory cost, keep the slice before quantization to avoid quantizing elements that will be discarded. Add coverage for the half-sized-slice boundary.
Differential Revision: D119550687