Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / nunoplopes/torchy issues and pull requests
#13 - Globals destruction order may cause deadlock with ~Trace
Issue -
State: open - Opened by nunoplopes over 2 years ago
#12 - hash collisions in program cache don't seem relevant
Issue -
State: open - Opened by nunoplopes about 3 years ago
#11 - Detection of dead ops breaks when ops return aliased storage
Issue -
State: open - Opened by nunoplopes about 3 years ago
#10 - Tensor reference in trace input leads to copies in make_variable
Issue -
State: open - Opened by nunoplopes about 3 years ago
- 1 comment
#9 - Trace cache: add shape, dtype, etc for key equality (but not the hash)
Issue -
State: open - Opened by nunoplopes about 3 years ago
#8 - isinstanceof(t, *Tensor) broken
Issue -
State: closed - Opened by nunoplopes about 3 years ago
- 2 comments
#7 - PyTorch's dispatcher callback mechanism seems broken
Issue -
State: closed - Opened by nunoplopes about 3 years ago
- 1 comment
#6 - PyTorch's ParsedArgs::tensor() copies the returned tensor
Issue -
State: open - Opened by nunoplopes about 3 years ago
#5 - arange.start_out: bad shape data
Issue -
State: open - Opened by nunoplopes about 3 years ago
#4 - shape inference doesn't use multiple Scalars
Issue -
State: open - Opened by nunoplopes about 3 years ago
#3 - super slow (10+ hours) shape inference
Issue -
State: open - Opened by nunoplopes about 3 years ago
#2 - Check why shallow_copy_from is called on a wrong object
Issue -
State: open - Opened by nunoplopes over 3 years ago
- 1 comment
#1 - PyTorch's redispatch functions copy input arguments
Issue -
State: closed - Opened by nunoplopes over 3 years ago
- 2 comments
Labels: performance