← back to Exo

fix m2 ultra flops

1fcbe18baa860ef35ecfe27d77b2ae2056cdb664 · 2024-07-20 21:37:27 -0700 · Alex Cheema

Files touched

• M exo/topology/device_capabilities.py

Diff

commit 1fcbe18baa860ef35ecfe27d77b2ae2056cdb664
Author: Alex Cheema <alexcheema123@gmail.com>
Date:   Sat Jul 20 21:37:27 2024 -0700

    fix m2 ultra flops
---
 exo/topology/device_capabilities.py | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/exo/topology/device_capabilities.py b/exo/topology/device_capabilities.py
index 5b681584..625384d3 100644
--- a/exo/topology/device_capabilities.py
+++ b/exo/topology/device_capabilities.py
@@ -47,13 +47,13 @@ CHIP_FLOPS = {
     # Note: currently no distinction between variants of M3 Max and M3 Pro, we pick the lower one to be conservative
     ### M chips
     "Apple M1": DeviceFlops(fp32=2.29*TFLOPS, fp16=4.58*TFLOPS, int8=9.16*TFLOPS),
-    "Apple M1 Max": DeviceFlops(fp32=10.60*TFLOPS, fp16=21.20*TFLOPS, int8=42.40*TFLOPS),
     "Apple M1 Pro": DeviceFlops(fp32=5.30*TFLOPS, fp16=10.60*TFLOPS, int8=21.20*TFLOPS),
+    "Apple M1 Max": DeviceFlops(fp32=10.60*TFLOPS, fp16=21.20*TFLOPS, int8=42.40*TFLOPS),
     "Apple M1 Ultra": DeviceFlops(fp32=21.20*TFLOPS, fp16=42.40*TFLOPS, int8=84.80*TFLOPS),
     "Apple M2": DeviceFlops(fp32=3.55*TFLOPS, fp16=7.10*TFLOPS, int8=14.20*TFLOPS),
-    "Apple M2 Max": DeviceFlops(fp32=13.49*TFLOPS, fp16=26.98*TFLOPS, int8=53.96*TFLOPS),
     "Apple M2 Pro": DeviceFlops(fp32=5.68*TFLOPS, fp16=11.36*TFLOPS, int8=22.72*TFLOPS),
-    "Apple M2 Ultra": DeviceFlops(fp32=10.60*TFLOPS, fp16=21.30*TFLOPS, int8=42.60*TFLOPS),
+    "Apple M2 Max": DeviceFlops(fp32=13.49*TFLOPS, fp16=26.98*TFLOPS, int8=53.96*TFLOPS),
+    "Apple M2 Ultra": DeviceFlops(fp32=26.98*TFLOPS, fp16=53.96*TFLOPS, int8=107.92*TFLOPS),
     "Apple M3": DeviceFlops(fp32=3.55*TFLOPS, fp16=7.10*TFLOPS, int8=14.20*TFLOPS),
     "Apple M3 Max": DeviceFlops(fp32=14.20*TFLOPS, fp16=28.40*TFLOPS, int8=56.80*TFLOPS),
     "Apple M3 Pro": DeviceFlops(fp32=4.97*TFLOPS, fp16=9.94*TFLOPS, int8=19.88*TFLOPS),

← 9d9d257e reduce chatgpt api response timeout in test  ·  back to Exo  ·  implement dynamic inference engine selection e9346641 →