Lyken17 edited a comment on pull request #10439:
URL: https://github.com/apache/tvm/pull/10439#issuecomment-1056647885
Thanks for the explaination. I agree it is important to annote the return
dtype in gradient calculation. But such information is missing when loading
models using `relay.fronted.from_xxx`. For example,
```python
import numpy as np
import torch
import torch.nn as nn
import tvm
from tvm import relay
from tvm.contrib import graph_executor
net = nn.Sequential(
nn.Conv2d(3, 3, 3, padding=1, groups=3)
)
input_shape = [1, 3, 32, 32]
input_data = torch.randn(input_shape)
input_name = "input0"
shape_list = [(input_name, input_data.shape)]
scripted_model = torch.jit.trace(net, input_data).eval()
fmod, params = relay.frontend.from_pytorch(scripted_model, shape_list,
default_dtype="float32")
mod = relay.transform.InferType()(fmod)
bwd_expr = relay.transform.gradient(mod["main"], mode="first_order")
bwd_mod = tvm.IRModule.from_expr(bwd_expr)
bwd_mod = relay.transform.InferType()(bwd_mod)
```
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
