issues
search
Lightning-AI
/
lightning-thunder
Make PyTorch models up to 40% faster! Thunder is a source to source compiler for PyTorch. It enables using different hardware executors at once; across one or thousands of GPUs.
Apache License 2.0
1.07k
stars
60
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Disallow in-place to view tensors
#630
crcrpar
closed
1 week ago
0
normalize test fix
#629
k223kim
closed
1 week ago
0
Make epilogue bits more uniform and useful
#628
t-vi
opened
1 week ago
0
Fix issue template
#627
t-vi
closed
1 week ago
0
recursive error fix
#626
k223kim
closed
5 days ago
3
cuDNN executor: No valid engine configs for MUL_Reduction_MUL_Matmul_MUL_ADD_SUB_EXP_Reshape_Matmul_Matmul_MUL_SUB_MUL_Reduction_MUL_Reshape_Matmul_Reshape_Matmul_
#625
t-vi
closed
1 week ago
2
[inplace][distributed] support `torch.distributed` common ops with `async_op=False`
#624
crcrpar
opened
1 week ago
0
support returning dataclasses with tensors
#623
t-vi
closed
4 days ago
0
Handles non-contiguous input strides
#622
vedaanta
closed
2 days ago
5
use `torch.get_default_dtype` and `torch.get_default_device` for factory method in `thunder/torch/__init__.py`
#621
jjsjann123
opened
1 week ago
2
Update error messages.
#620
tfogal
closed
1 week ago
0
`saved_tensors_list` might contain `None`s
#619
nikitaved
opened
1 week ago
0
[feature request] Numpy 2 compatibility
#618
nikitaved
opened
1 week ago
1
fix numpy version for tests
#617
nikitaved
closed
1 week ago
0
test: no Nones in saved_tensors_list
#616
nikitaved
opened
1 week ago
1
Match torch signature for `torch.nn.functional.silu`
#615
riccardofelluga
closed
1 week ago
0
[Feature request] Optional debugging option to get trace with information on tensor strides along with tensor shapes
#614
parthmannan
opened
1 week ago
3
Allowing static constraint in torch/__init__.py
#613
jjsjann123
closed
5 days ago
2
[pre-commit.ci] pre-commit suggestions
#612
pre-commit-ci[bot]
closed
1 week ago
1
`numpy` version requirement fix
#611
nikitaved
closed
1 week ago
1
numpy 2.0.0 dtypes changes
#610
crcrpar
closed
1 week ago
2
NumPy 2.0 compat: replace np.float_ with np.float64
#609
nikitaved
closed
1 week ago
1
Fix flaky torch max grad
#608
kshitij12345
closed
1 week ago
2
replace clear_mutable_collection with a data structure
#607
nikitaved
closed
2 days ago
1
Revise memory clearing mechanism in the torch.autograd.Function integration
#606
IvanYashchuk
opened
1 week ago
6
Add nvFuser support for thunder.torch.randn
#605
IvanYashchuk
opened
1 week ago
1
thunder.torch.randn should accept any Sequence of integers as a shape argument
#604
IvanYashchuk
opened
1 week ago
0
jit error: unpacking from nonconstant opaque function
#603
IvanYashchuk
opened
1 week ago
4
fsdp(jit(model)) + parameter sharing - dont duplicate allgather
#602
kshitij12345
closed
1 week ago
1
unhashable type: 'TensorProxy' error on NeVA model
#601
tfogal
closed
1 day ago
2
Add a transform to add nvtx range on the optimized trace
#600
kshitij12345
closed
1 week ago
3
Add torch.max
#599
kshitij12345
closed
1 week ago
1
use provided comp in test_grad.py
#598
k223kim
closed
1 week ago
1
Partially support in-place ops and tensor aliases
#597
crcrpar
closed
5 days ago
9
Restore usage of CUDAGraphExecutor with compiled backward function
#596
IvanYashchuk
closed
2 weeks ago
2
Remove _create_callable usage
#595
IvanYashchuk
closed
1 day ago
5
[Tensor Parallelism] Improve comm optimization logic for pair of column-wise parallel linear and row-wise parallel linear
#594
crcrpar
opened
2 weeks ago
0
Bumps cudnn FE to v1.5
#593
vedaanta
closed
1 week ago
6
Sort allgathers according to consumer order, reduce scatter according to producer order
#592
kiya00
closed
2 days ago
5
TE: fix the placement of bwd_sync symbol in trace
#591
kshitij12345
closed
2 weeks ago
0
more in-place ops
#590
crcrpar
closed
2 weeks ago
0
Warn when torch._C._set_grad_enabled is called
#589
kshitij12345
closed
2 weeks ago
0
An error occurred: KeyError – 't5479' /
#588
wprazuch
closed
1 week ago
5
Unsupported – setattr(FSDPManagedNNModuleVariable(FullyShardedDataParallel), _is_root, ...)
#587
wprazuch
closed
1 week ago
2
adding Static Constraint field in NumberProxy
#586
jjsjann123
closed
1 week ago
3
Copy to the original tensors
#585
crcrpar
closed
2 weeks ago
0
Functionalize in-place ops
#584
crcrpar
closed
1 week ago
13
CUDA error: CUDA_ERROR_ILLEGAL_ADDRESS failed when training falcon-7b
#583
mpatel31415
closed
2 days ago
11
NotImplementedError: requires_grad=True is not yet supported within thunder.compile
#582
mpatel31415
opened
2 weeks ago
7
nvFuser executor doesn't support prims.sum with symbolic dimensions
#581
IvanYashchuk
opened
2 weeks ago
5
Previous
Next