Comments (3)
fyi @jjsjann123 This one could be tricky
from lightning-thunder.
fyi @jjsjann123 This one could be tricky
oh yes, we don't have control flows and this check could happen somewhere in the middle that'll be hard to handle...
Does forcing the branch
sounds like a reasonable~ish short-term solution to you @tfogal ? We can get the function up first and then think about how do we faithfully execute user program then.
is_contiguous + true_block/false_block
-> contiguous + true_block
from lightning-thunder.
triage review:
- modeling
t.is_contiguous
asreturn True
sounds like a reasonable workaround for now, but there are some issues... - this would require updating the program to reuse the output of the contiguous call as the the input to the
is_contiguous
query - (we should check that
.contiguous()
creates a new tensor) - in the future maybe we should model memory format contiguous or a particular stride order
- we should consider what happens if
t
is just one name for the tensor it refers to in the program
from lightning-thunder.
Related Issues (20)
- Support NeMo NeVa Model
- Support NeMo MegatronGPTModel
- import thunder tries to initialize all native executors HOT 3
- Support `torch.nn.functional.adaptive_avg_pool2d` HOT 1
- `use_cudagraphs=True` is silently ignored. HOT 1
- Distributed and Bucketing Performance Improvements HOT 1
- using size() twice in a row on same tensor causes interpreter out of sync error with thunder jit HOT 1
- Consider if functions like full_like and expand_as should be symbols (or not)
- max_pool2d:Shape propagation error for input dimensions used by resnet HOT 1
- transformer_engine: clear_collection doesn't clear the tensors saved by transformer_engine_ex ops for backward. HOT 1
- transformer_engine: Fails with Pythia-14m in distributed
- Benchmark: Running benchmark_litgpt with `transformerengine` leads to `nullcontext` not defined.
- uniform_like: The function outputs different values when the input tensor is the same but `requires_grad` is True/False.
- Add activation checkpointing to litGPT benchmarking script HOT 1
- utils.find_producer_symbols is too aggressive at collecting producer symbols not taking into account stop information HOT 1
- Add more nvFuser debug information HOT 2
- reenable apex backward / phantom grad tests without wrecking CI HOT 2
- Investigate GPU memory intense CI tests HOT 1
- Regression on thunder/benchmarks/targets.py
- CSE pass should work on symbols with keyword arguments
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 lightning-thunder.