Comments (2)
This seems related to #122243
The documentation for the log_softmax
argument does explicitly state that it should be obtained from the output of torch.nn.functional.log_softmax()
, so this behavior seems to be expected
We are happy to accept a PR that addresses this issue however
from pytorch.
Closing duplicate in favor of #52241
from pytorch.
Related Issues (20)
- Segmentaion fault on Apple M1 caused by libomp on v2.4.1 HOT 6
- [pipelining] create two stage instances out of same stage module HOT 1
- [pipelining] try not to dry run module when creating PipelineStage? HOT 3
- [Inductor] Some mechanism to fuse pointwise ops (and more) into a user-defined triton kernel
- Can not load full state dictionary containing parameters whose first dimension is not divisible by DP size in FSDP 2 + TP model HOT 2
- DISABLED test_mem_leak_from_save_for_bw (__main__.TestAOTAutogradWithDynamo) HOT 1
- DISABLED test_graph_grad_scaling_foreach_False_fused_True_AdamW_cuda_float32 (__main__.TestCudaOptimsCUDA) HOT 2
- High MacOS queue HOT 11
- NaN in Flex Attention backward if BlockMask is larger than first run seq_len when return_lse=True and torch compiled
- profiler for PT2 can give wrong compilation frame ID
- reflectionPad2d cuda(and CPU?) might have incorrect results - need clarification. HOT 2
- Async NCCL communication blocks CUDA kernel in the first run
- A question about `getDLDevice`
- Performance regression in torch.compile
- TorchScript raises an OSError when using frozen dataclass HOT 1
- torch.compile doesn't generate any graphs when modules are patched with new parameters class HOT 1
- `to()` Method Does Not Move Internal Components of Sparse Tensors
- When calculating the loss, the input data does not contain NaN, but the output contains NaN.
- Inductor generates sync copy_() on .to(device, non_blocking=False)
- Flex Attention Extremely Slow HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from pytorch.