You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
"allow_gpu_fallback": false, # (DLA only) Allow layers unsupported on DLA to run on GPU
238
259
},
239
260
"op_precision": torch.half, # Operating precision set to FP16
261
+
# List of datatypes that should be configured for each input. Supported options torch.{float|half|int8|int32|bool}.
240
262
"disable_tf32": False, # Force FP32 layers to use traditional as FP32 format vs the default behavior of rounding the inputs to 10-bit mantissas before multiplying, but accumulates the sum using 23-bit mantissas
"allow_gpu_fallback": false, # (DLA only) Allow layers unsupported on DLA to run on GPU
42
42
},
43
43
"op_precision": torch.half, # Operating precision set to FP16
44
+
"input_dtypes": [torch.float32] # List of datatypes that should be configured for each input. Supported options torch.{float|half|int8|int32|bool}.
44
45
"refit": false, # enable refit
45
46
"debug": false, # enable debuggable engine
46
47
"strict_types": false, # kernels should strictly run in operating precision
@@ -106,6 +107,7 @@ def convert_method_to_trt_engine(module: torch.jit.ScriptModule, method_name: st
106
107
"allow_gpu_fallback": false, # (DLA only) Allow layers unsupported on DLA to run on GPU
107
108
},
108
109
"op_precision": torch.half, # Operating precision set to FP16
110
+
# List of datatypes that should be configured for each input. Supported options torch.{float|half|int8|int32|bool}.
109
111
"disable_tf32": False, # Force FP32 layers to use traditional as FP32 format vs the default behavior of rounding the inputs to 10-bit mantissas before multiplying, but accumulates the sum using 23-bit mantissas
0 commit comments