You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by "MasterJH5574 (via GitHub)" <gi...@apache.org> on 2023/03/21 02:59:41 UTC
[GitHub] [tvm] MasterJH5574 commented on a diff in pull request #14348: [Unity] Add More Ops For FX Translator
MasterJH5574 commented on code in PR #14348:
URL: https://github.com/apache/tvm/pull/14348#discussion_r1142847893
##########
tests/python/relax/test_frontend_from_fx.py:
##########
@@ -19,20 +19,51 @@
import tvm
from tvm import relax
import tvm.testing
-from tvm.script.parser import ir as I, relax as R, tir as T
+from tvm.script import ir as I
+from tvm.script import relax as R
+from tvm.script import tir as T
-def verify_model(torch_model, input_info, binding, expected):
+def verify_model(torch_model, input_info, binding, expected, use_dynamo=False):
+ import torch
+ import torch._dynamo as dynamo
from torch import fx
from tvm.relax.frontend.torch import from_fx
- graph_model = fx.symbolic_trace(torch_model)
- mod = from_fx(graph_model, input_info)
+ if use_dynamo:
+ args = []
+ for info in input_info:
+ args.append(torch.zeros(*info[0], dtype=_convert_data_type(info[1])))
+ graph_model = dynamo.export(torch_model, *args)[0]
+ else:
+ graph_model = fx.symbolic_trace(torch_model)
+ print(graph_model.code)
+ mod = from_fx(graph_model, input_info, unwrap_unit_return_tuple=True)
Review Comment:
For `unwrap_unit_return_tuple=True` I think it depends. Usually for graph from `dynamo.export`, you want to use `unwrap_unit_return_tuple=True`. But for graphs from `fx.symbolic_trace`, it is not necessary to use True. We also want to test the case of `unwrap_unit_return_tuple=False` in this file. Seems that with this change it is untested.
##########
tests/python/relax/test_frontend_from_fx.py:
##########
@@ -19,20 +19,51 @@
import tvm
from tvm import relax
import tvm.testing
-from tvm.script.parser import ir as I, relax as R, tir as T
+from tvm.script import ir as I
+from tvm.script import relax as R
+from tvm.script import tir as T
-def verify_model(torch_model, input_info, binding, expected):
+def verify_model(torch_model, input_info, binding, expected, use_dynamo=False):
+ import torch
+ import torch._dynamo as dynamo
from torch import fx
from tvm.relax.frontend.torch import from_fx
- graph_model = fx.symbolic_trace(torch_model)
- mod = from_fx(graph_model, input_info)
+ if use_dynamo:
+ args = []
+ for info in input_info:
+ args.append(torch.zeros(*info[0], dtype=_convert_data_type(info[1])))
+ graph_model = dynamo.export(torch_model, *args)[0]
+ else:
+ graph_model = fx.symbolic_trace(torch_model)
+ print(graph_model.code)
+ mod = from_fx(graph_model, input_info, unwrap_unit_return_tuple=True)
+ print(mod.script())
Review Comment:
Debugging lines to remove.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org