mlboydaisuke commited on
Commit
1db68c2
·
verified ·
1 Parent(s): fbf5773

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +6 -0
README.md CHANGED
@@ -29,6 +29,12 @@ All variants take and return fp32 tensors — swap the `.pte` file, keep your ap
29
  \*Mac arm64, single process, median of 10 — a reference point for relative cost
30
  only, not a device number (torch eager fp32 on the same machine: 375.5 ms).
31
 
 
 
 
 
 
 
32
  ### Precisions that did not earn a slot
33
 
34
  - **fp16 is not shipped**: it comes out at 100% of the fp32 file (176.1 MB vs 176.1 MB), so it buys nothing. XNNPACK serializes convolution weights as fp32 no matter what dtype the graph carries, so on a conv-heavy model fp16 saves no disk and only adds cast operations. Reach for int8 here, not fp16.
 
29
  \*Mac arm64, single process, median of 10 — a reference point for relative cost
30
  only, not a device number (torch eager fp32 on the same machine: 375.5 ms).
31
 
32
+ ### Checked in the task's own units
33
+
34
+ Correlation is a first filter. These are the numbers that decide:
35
+
36
+ - **int8** — measured in the units that matter for this model: mask IoU at 0.5, median 0.9994 over five real images (worst 0.9916) against the fp32 build.
37
+
38
  ### Precisions that did not earn a slot
39
 
40
  - **fp16 is not shipped**: it comes out at 100% of the fp32 file (176.1 MB vs 176.1 MB), so it buys nothing. XNNPACK serializes convolution weights as fp32 no matter what dtype the graph carries, so on a conv-heavy model fp16 saves no disk and only adds cast operations. Reach for int8 here, not fp16.