Changes done internally at Facebook

364639a8ab2ee7531ce5259b8985a3c90bda4fdf Wei Wei <[email protected]> [fx2trt] target files added
07d8e842b54b9c727f4215239f6c007cc7a62c9f Wei Wei <[email protected]> Swap fx2trt_oss to torch_tensorrt
74731c90fd63e41ff5997887d8f72ca0b805cf8d Yinghai Lu <[email protected]> Fix uru_10x10 test
6c53d36a08a7d465a1108d7154ef29a373eb38cc Wei Wei <[email protected]> [fx2trt] Modify lower setting class to accommandate AIT lowering
6f873f4f3ece9d476479eb7c9633d38554dd8692 Oleg Khabinov <[email protected]> [fx2trt] Make sure acc_tracer belongs only to single target
529a5750ace2bede6e9b7a9922a0f75c459df16b Shirong Wu <[email protected]> Enable explicit batch dim for MTS gpu benchmark
2d284df94ddb530f3a8875fdc76796fad508ec29 Wei Wei <[email protected]> [fx2trt] remove wildcard for obj of torch_fx2trt in TARGETS
84b53b15427cc08fb1e36143b6bdec4557f50d7e Shirong Wu <[email protected]> Add var converter
17e309b17b3ba66cda0e7d5712089d860a5e125e Jordan Fix <[email protected]> [const_fold] Set requires_grad based on the folded tensor; add device_for_folding option
2c8f1b23be30ec968ad27215256d250c872616b0 Kefei Lu <[email protected]> lowering: support creating lowerer instance with "presets"
fe350dcc4d6ff19c76191fcf3690349eaf9dfaa7 Oleg Khabinov <[email protected]> [fx_acc] Add acc_tracer support for torch.mm

Author: khabinov

docs/v1.1.0/._index.html

@@ -739,6 +739,13 @@ def square_mapper(node: torch.fx.Node, _: nn.Module) -> torch.fx.Node:
         ("mat2", "other"),
     ],
 )
+@register_acc_op_mapping(
+    op_and_target=("call_function", torch.mm),
+    arg_replacement_tuples=[
+        ("input", "input"),
+        ("mat2", "other"),
+    ],
+)
 @register_acc_op_mapping(op_and_target=("call_function", torch.matmul))
 @register_acc_op
 def matmul(*, input, other):