File tree Expand file tree Collapse file tree 1 file changed +13
-1
lines changed Expand file tree Collapse file tree 1 file changed +13
-1
lines changed Original file line number Diff line number Diff line change @@ -1042,8 +1042,20 @@ def __len__(self):
1042
1042
1043
1043
@functools .lru_cache (None )
1044
1044
def is_big_gpu (index ) -> bool :
1045
+ prop = torch .cuda .get_device_properties (index )
1046
+
1047
+ # SM logic is not relevant to ROCm gpus
1048
+ # Arbitrarily skipping the older models
1049
+ if torch .version .hip is not None :
1050
+ if prop .major < 9 or prop .major == 10 :
1051
+ log .warning (
1052
+ "GPU arch does not support max_autotune_gemm mode usage"
1053
+ )
1054
+ return False
1055
+ return True
1056
+
1045
1057
min_sms = 68 # 3080
1046
- avail_sms = torch . cuda . get_device_properties ( index ) .multi_processor_count
1058
+ avail_sms = prop .multi_processor_count
1047
1059
if avail_sms < min_sms :
1048
1060
log .warning (
1049
1061
"Not enough SMs to use max_autotune_gemm mode" ,
You can’t perform that action at this time.
0 commit comments