gmagogsfm

Results 10 comments of gmagogsfm

I tried to write an example with Identity module at top of trunk PyTorch, the issue is not reproducible. Could you try loading a toy module generated by this script...

This is the serialized code when compiled in my environment: ``` def forward(self: __torch__.torch.nn.modules.linear.Identity, input: Tensor) -> Tensor: ``` As you can see it is different from the one shown...

> > I tried to write an example with Identity module at top of trunk PyTorch, the issue is not reproducible. > > Could you try loading a toy module...

Reed, thanks for bringing this to my attention. @mrmhodak For the GRPC error. Can it be reproduced stably? I am asking because XLA only really starts affecting your computation when...

> > Is the import process finished? If not, should we manually merge this PR again? > > @thiagocrepaldi I'm seeing tons of executorch errors internally, need to figure that...

@zou3519 This is one of the smaller bugs I found while investigating #28077. Please take a look

@zou3519 @ProExpertProg Could you take a quick look at this one-line change? Thanks

> @gmagogsfm could you just elaborate what the repro for causing this issue is? I discovered this issue when debugging a dynamic shape bug. I think `inp` was set to...

I did some investigations, here are my findings and suggestions. The main problem in this bug that when `max_num_batched_tokens` is set to 1, this forces `input_ids` to be shape `(1,)`....