Summary:
`torch.linspace(0, 1, 1)` fails with `RuntimeError: invalid argument 3: invalid number of points at ../aten/src/TH/generic/THTensorMoreMath.cpp:2119`, while `np.linspace(0, 1, 1)` works fine.
Looking at the code, there is even a comment by gchanan asking: "NumPy allows you to pass different points even if n <= 1 -- should we?"
I would say "yes". Currently, I would need to handle the case of `steps == 1` or `steps == 0` separately, making sure to change the `end` when calling `torch.linspace`. This is impractical. If we support `start != end`, there are two possibilities for the result: Either we ensure the first value in the resulting sequence always equals `start`, or we ensure the last value in the resulting sequence always equals `end`. Numpy chose the former, which also allows it to support a boolean `endpoint` flag. I'd say we should follow numpy.
This PR adapts `linspace` and `logspace` to mimic the behavior of numpy, adapts the tests accordingly, and extends the docstrings to make clear what happens when passing `steps=1`.
If you decide against this PR, the error message should become explicit about what I did wrong, and the documentation should be extended to mention this restriction.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14748
Differential Revision: D13356136
Pulled By: ezyang
fbshipit-source-id: db85b8f0a98a5e24b3acd766132ab71c91794a82
Summary:
Before this PR, tensor.clamp() would return an empty tensor if min and
max were not specified. This is a regression from 0.4.1, which would
throw an error. This PR restores that error message.
Fixes#14470
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14716
Differential Revision: D13311031
Pulled By: zou3519
fbshipit-source-id: 87894db582d5749eaccfc22ba06aac4e10983880
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13603
P
Moved vectorized CPU copy to aten. Notable changes mainly in _copy_same_type_.
Reviewed By: ezyang
Differential Revision: D12936031
fbshipit-source-id: 00d28813e3160595e73d104f76685e13154971c1
Summary:
Multi-dimensional `sum` is already implemented, and it's trivial to implement `mean` in terms of `sum`, so just do it.
Bonus: Fix incomplete language in the `torch.sum` documentation which doesn't take into account multiple dimensions when describing `unsqueeze` (at the same time as introducing similar language in `torch.mean`).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14252
Differential Revision: D13161157
Pulled By: umanwizard
fbshipit-source-id: c45da692ba83c0ec80815200c5543302128da75c
Summary:
Fixes https://github.com/pytorch/pytorch/issues/14344 and https://github.com/pytorch/pytorch/issues/6863
The slowdown was due to the fact that we were only summarizing the tensor (for computing the number of digits to print) if its first dimension was larger than the threshold. It now goes over all the dimensions.
Some quick runtime analysis:
Before this PR:
```python
In [1]: import torch; a = torch.rand(1, 1700, 34, 50)
In [2]: %timeit str(a)
13.6 s ± 84.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
```
After this PR
```python
In [1]: import torch; a = torch.rand(1, 1700, 34, 50)
In [2]: %timeit str(a)
2.08 ms ± 395 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
In [3]: b = a.cuda()
In [4]: %timeit str(b)
8.39 ms ± 45.9 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14418
Reviewed By: weiyangfb
Differential Revision: D13226950
Pulled By: soumith
fbshipit-source-id: 19eb4b855db4c8f891d0925a9c56ae8a2824bb23
Summary:
They didn't turn up in my tests because I use pytest which doesn't
print debug statements if the tests pass
Differential Revision: D13115227
Pulled By: soumith
fbshipit-source-id: 46a7d47da7412d6b071158a23ab21e7fb0c6e11b
Summary:
Implements batching for the Cholesky decomposition.
Performance could be improved with a dedicated batched `tril` and `triu` op, which is also impeding autograd operations.
Changes made:
- batching code
- tests in `test_torch.py`, `test_cuda.py` and `test_autograd.py`.
- doc string modification
- autograd modification
- removal of `_batch_potrf` in `MultivariateNormal`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14017
Differential Revision: D13087945
Pulled By: ezyang
fbshipit-source-id: 2386db887140295475ffc247742d5e9562a42f6e
Summary:
This enables the distributions and utils test sets for ROCm.
Individual tests are enabled that now pass due to fixes in HIP/HCC/libraries versions in white rabbit.
For attention: bddppq ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13166
Differential Revision: D12814759
Pulled By: bddppq
fbshipit-source-id: ea70e775c707d7a8d2776fede6154a755adef43e
Summary:
- This is a straightforward PR, building up on the batch inverse PR, except for one change:
- The GENERATE_LINALG_HELPER_n_ARGS macro has been removed, since it is not very general and the resulting code is actually not very copy-pasty.
Billing of changes:
- Add batching for `potrs`
- Add relevant tests
- Modify doc string
Minor changes:
- Remove `_gesv_single`, `_getri_single` from `aten_interned_strings.h`.
- Add test for CUDA `potrs` (2D Tensor op)
- Move the batched shape checking to `LinearAlgebraUtils.h`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13453
Reviewed By: soumith
Differential Revision: D12942039
Pulled By: zou3519
fbshipit-source-id: 1b8007f00218e61593fc415865b51c1dac0b6a35
Summary:
update roll to behave as in numpy.roll when dimension to roll not specified.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13588
Differential Revision: D12964295
Pulled By: nairbv
fbshipit-source-id: de9cdea1a937773033f081f8c1505a40e4e08bc1
Summary:
- a walk around for #13292, a complete fix requires investigation on the root cause when using advanced indexing
- this PR brings in `filp()` CUDA implementation for CPU kernel
- with this change:
```
>>> t = torch.randn(1, 3, 4, 5)
>> t.flip(1, 3).shape
torch.Size([1, 3, 4, 5])
```
- performance:
```
====== with this PR ======
>>> a = torch.randn(1000, 1000)
>>> %timeit -r 100 a.flip(0, 1)
1.98 ms ± 579 µs per loop (mean ± std. dev. of 100 runs, 1000 loops each)
====== Perf at previous PR #7873 ======
100 loops, best of 3: 11 ms per loop
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13344
Differential Revision: D12968003
Pulled By: weiyangfb
fbshipit-source-id: 66f434049d143a0575a35b5c983b3e0577a1a28d
Summary:
- fixes weights-contiguous requirement for THCUNN Convolutions
- Add tests that conv backward pass works for non-contiguous weights
- fix RNN tests / error messages to be consistent and pass
- relax weight grad precision for fp16 for a particular test
- fix regression of CMAKE_PREFIX_PATH not passing through
- add missing skipIfNoLapack annotations where needed
Differential Revision: D12918456
Pulled By: soumith
fbshipit-source-id: 8642d36bffcc6f2957800d6afa1e10bef2a91d05
Summary:
Fixes#13326
Also now you can use `run_test.py` with `pytest`. E.g.,
```
python run_test.py -vci distributed -pt
```
Yes it works with `distributed` and `cpp_extension`.
cc zou3519 vishwakftw
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13416
Differential Revision: D12895622
Pulled By: SsnL
fbshipit-source-id: 2d18106f3a118d642a666bfb1318f41c859c3df7
Summary:
This PR performs a renaming of the function `potrf` responsible for the Cholesky
decomposition on positive definite matrices to `cholesky` as NumPy and TF do.
Billing of changes
- make potrf cname for cholesky in Declarations.cwrap
- modify the function names in ATen/core
- modify the function names in Python frontend
- issue warnings when potrf is called to notify users of the change
Reviewed By: soumith
Differential Revision: D10528361
Pulled By: zou3519
fbshipit-source-id: 19d9bcf8ffb38def698ae5acf30743884dda0d88
Summary:
Currently, `a = 1 - torch.tensor([1]).to('cuda:1')` puts `a` in `cuda:1` but reports `a.device` as `cuda:0` which is incorrect, and it causes illegal memory access error when trying to access `a`'s memory (e.g. when printing). This PR fixes the error.
Fixes https://github.com/pytorch/pytorch/issues/10850.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12956
Differential Revision: D12835992
Pulled By: yf225
fbshipit-source-id: 5737703d2012b14fd00a71dafeedebd8230a0b04
Summary:
ezyang on the template hack
smessmer on SFINAE of the `TensorOptions(Device)`
goldsborough on the C++ API test changes
zdevito on the `jit` codegen changes
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13146
Reviewed By: ezyang
Differential Revision: D12823809
Pulled By: SsnL
fbshipit-source-id: 98d65c401c98fda1c6fa358e4538f86c6495abdc
Summary:
1. Refactors `TestTorch` into `TestTorchMixin` (subclass of `object`) and `TestTorch` (subclass of `TestCase`, MRO `(TestCase, TestTorchMixin)`, only defined if `__name__ == '__main__'`). So other scripts won't accidentally run it.
2. Adds an assertion in `load_tests` that each script only runs cases defined in itself.
cc yf225 ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13250
Differential Revision: D12823734
Pulled By: SsnL
fbshipit-source-id: 7a169f35fe0794ce76e310d8a137d9a3265c012b
Summary:
Fixes#12578#9395.
* Fix and simplify print logic
* Follow numpy print rule eb2bd11870/numpy/core/arrayprint.py (L859)
> scientific notation is used when absolute value of the smallest number is < 1e-4 or maximum > 1e8 or the ratio of the maximum absolute value to the minimum is > 1e3
I hope I didn't break anything since there seems to be a lot of edge cases here... Here are some easy sanity checks.
```
In [5]: torch.tensor(1)
Out[5]: tensor(1)
Out[2]: array(1) # numpy
In [6]: torch.tensor(10)
Out[6]: tensor(10)
Out[3]: array(10) # numpy
In [8]: torch.tensor(99000000)
Out[8]: tensor(99000000)
Out[5]: array(99000000) # numpy
In [9]: torch.tensor(100000000)
Out[9]: tensor(100000000)
Out[6]: array(100000000) # numpy
In [10]: torch.tensor(100000001)
Out[10]: tensor(100000001)
Out[7]: array(100000001) # numpy
In [11]: torch.tensor(1000000000)
Out[11]: tensor(1000000000)
Out[8]: array(1000000000) # numpy
In [12]: torch.tensor([1, 1000])
Out[12]: tensor([ 1, 1000])
Out[9]: array([ 1, 1000]) # numpy
In [13]: torch.tensor([1, 1010])
Out[13]: tensor([ 1, 1010])
Out[10]: array([ 1, 1010]) # numpy
```
For floating points, we use scientific when `max/min > 1000 || max > 1e8 || min < 1e-4`
Lines with "old" are old behaviors that either has precision issue, or not aligned with numpy
```
In [14]: torch.tensor(0.01)
Out[14]: tensor(0.0100)
Out[11]: array(0.01) # numpy
In [15]: torch.tensor(0.1)
Out[15]: tensor(0.1000)
Out[12]: array(0.1) # numpy
In [16]: torch.tensor(0.0001)
Out[16]: tensor(0.0001)
Out[14]: array(0.0001) # numpy
In [17]: torch.tensor(0.00002)
Out[17]: tensor(2.0000e-05)
Out[15]: array(2e-05) # numpy
Out[5]: tensor(0.0000) # old
In [18]: torch.tensor(1e8)
Out[18]: tensor(100000000.)
Out[16]: array(100000000.0) # numpy
In [19]: torch.tensor(1.1e8)
Out[19]: tensor(1.1000e+08)
Out[17]: array(1.1e8) # numpy 1.14.5, In <= 1.13 this was not using scientific print
Out[10]: tensor(110000000.) # old
In [20]: torch.tensor([0.01, 10.])
Out[20]: tensor([ 0.0100, 10.0000])
Out[18]: array([ 0.01, 10. ]) # numpy
In [21]: torch.tensor([0.01, 11.])
Out[21]: tensor([1.0000e-02, 1.1000e+01])
Out[19]: array([ 1.00000000e-02, 1.10000000e+01]) # numpy
Out[7]: tensor([ 0.0100, 11.0000]) # old
```
When print floating number in int mode, we still need to respect rules to use scientific mode first
```
In [22]: torch.tensor([1., 1000.])
Out[22]: tensor([ 1., 1000.])
Out[20]: array([ 1., 1000.]) # numpy
In [23]: torch.tensor([1., 1010.])
Out[23]: tensor([1.0000e+00, 1.0100e+03])
Out[21]: array([ 1.00000000e+00, 1.01000000e+03]) # numpy
Out[9]: tensor([ 1., 1010.]) # old
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12746
Differential Revision: D10443800
Pulled By: ailzhang
fbshipit-source-id: f5e4e3fe9bf0b44af2c64c93a9ed42b73fa613f5
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12794
common.py is used in base_module for almost all tests in test/. The
name of this file is so common that can easily conflict with other dependencies
if they happen to have another common.py in the base module. Rename the file to
avoid conflict.
Reviewed By: orionr
Differential Revision: D10438204
fbshipit-source-id: 6a996c14980722330be0a9fd3a54c20af4b3d380
Summary:
I found a bug in norm() and fixed it (and added tests to make sure it's fixed)
here is how to reproduce it:
```python
import torch
x = torch.FloatTensor([[10, 12, 13], [4, 0, 12]])
print(torch.norm(x, -40, dim=0, keepdim=True)) #output is tensor([[ 4.0000, 0.0000, 11.9853]])
print(torch.norm(x, float('-inf'), dim=0, keepdim=True)) #output is tensor([[1., 1., 1.]]) which is wrong!
from numpy.linalg import norm as np_norm
x = x.numpy()
print(np_norm(x, ord=-40, axis=0)) #output is array([[4., 0., 11.985261]])
print(np_norm(x, ord=float('-inf'), axis=0)) #output is array([[4., 0., 12.0]])
```
it's related to [#6817](https://github.com/pytorch/pytorch/issues/6817) and [#6969](https://github.com/pytorch/pytorch/pull/6969)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12722
Differential Revision: D10427687
Pulled By: soumith
fbshipit-source-id: 936a7491d1e2625410513ee9c39f8c910e8e6803
Summary:
`torch.isfinite()` used to crash on int inputs.
```
>>> import torch
>>> a = torch.tensor([1, 2])
>>> torch.isfinite(a)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/scratch/pytorch/torch/functional.py", line 262, in isfinite
return (tensor == tensor) & (tensor.abs() != inf)
RuntimeError: value cannot be converted to type int64_t without overflow: inf
```
But this is a easy special case and numpy also supports it.
```
>>> import numpy as np
>>> a = np.array([1, 2])
>>> a.dtype
dtype('int64')
>>> np.isfinite(a)
array([ True, True], dtype=bool)
```
So added a hacky line to handle non-floating-point input. Since pytorch raises exception when overflow, we can safely assume all valid int tensors are infinite numbers.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12750
Differential Revision: D10428204
Pulled By: ailzhang
fbshipit-source-id: f39b2d0975762c91cdea23c766ff1e21d85d57a5
Summary:
The mapping protocol stipulates that when `__delitem__` is called, this is passed to `__setitem__` [(well, the same function in the C extension interface)](https://docs.python.org/3/c-api/typeobj.html#c.PyMappingMethods.mp_ass_subscript) with NULL data.
PyTorch master crashes in this situation, with this patch, it does not anymore.
Test code (careful, sefaults your interpreter):
```python
import torch
a = torch.randn(5)
del a[2]
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12726
Differential Revision: D10414244
Pulled By: colesbury
fbshipit-source-id: c49716e1a0a3d9a117ce88fc394858f1df36ed79
Summary:
- This was one of the few functions left out from the list of functions in
NumPy's `linalg` module
- `multi_mm` is particularly useful for DL research, for quick analysis of
deep linear networks
- Added tests and doc string
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12380
Differential Revision: D10357136
Pulled By: SsnL
fbshipit-source-id: 52b44fa18d6409bdeb76cbbb164fe4e88224458e
Summary:
* switches docker files over to white rabbit release - removed custom package installs
* skips five tests that regressed in that release
* fixes some case-sensitivity issues in ROCm supplied cmake files by sed'ing them in the docker
* includes first changes to the infrastructure to support upcoming hip-clang compiler
* prints ROCm library versions as part of the build (as discussed w/ ezyang )
* explicitly searches for miopengemm
* installs the new hip-thrust package to be able to remove the explicit Thrust checkout in a future revision
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12577
Differential Revision: D10350165
Pulled By: bddppq
fbshipit-source-id: 60f9c9caf04a48cfa90f4c37e242d944a175ab31
Summary:
Fixes#12260#2896
```
torch.multinomial(torch.FloatTensor([0, 1, 0, 0]), 3, replacement=False)
```
The old behavior is that we return `0` after we run out of postive categories. Now we raise an error based on discussion in the issue thread.
- Add testcase for cpu & cuda case, in cuda case `n_samples=1` is a simple special case, so we test against `n_sample=2` instead.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12490
Differential Revision: D10278794
Pulled By: ailzhang
fbshipit-source-id: d04de7a60f60d0c0d648b975db3f3961fcf42db1
Summary:
* Topk part 1: fix intrinsincs for 64 wave front (#224)
64 in a wave front - intrinsics change.
* Disable in-place sorting on ROCm. (#237)
It is known to hang - use the Thrust fallback
Skip one test - fails with the fallback.
* Topk fixes (#239)
* Spec (https://docs.nvidia.com/cuda/pdf/ptx_isa_6.3.pdf) Sec 9.7.1.19 (bfe) and 9.7.1.20 (bfi) requires pos and len to be limited to 0...255
* Spec (https://docs.nvidia.com/cuda/pdf/ptx_isa_6.3.pdf) Sec 9.7.1.19 requires extracted bits to be in LSBs
* Correct logic for getLaneMaskLe. Previous logic would return 0x0 instead of 0xffffffffffffffff for lane 63
* Round up blockDim.x to prevent negative index for smem
bddppq ezyang
Note the one additional skipped test resulting from using the thrust sort fallback for all sizes. We are working on getting bitonic to work properly (and always). Until then, this needs to be skipped on ROCm.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12337
Differential Revision: D10259481
Pulled By: ezyang
fbshipit-source-id: 5c8dc6596d7a3103ba7b4b550cba895f38c8148e
Summary:
- fixes https://github.com/pytorch/pytorch/issues/10723
- migrate PReLU to ATen and deprecate legacy PReLU
- performance:
CPU with weight.numel() = 1
```
>>> m = nn.PReLU()
>>> x = torch.randn(100, 100, 100, requires_grad=True)
>>> %timeit -r 100 y = m(x)
100 loops, best of 100: 9.43 ms per loop
>>> y = m(x).sum()
>>> %timeit -r 100 y.backward(retain_graph=True)
10 loops, best of 100: 24.4 ms per loop
>>> m = nn.PReLU()
>>> x = torch.randn(100, 100, 100, requires_grad=True)
>>> %timeit -r 100 y = m(x)
1000 loops, best of 100: 695 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 y.backward(retain_graph=True)
100 loops, best of 100: 2.47 ms per loop
```
CPU with weight.numel() = channels
```
>>> m = nn.PReLU(100)
>>> x = torch.randn(100, 100, 100, requires_grad=True)
>>> %timeit -r 100 y = m(x)
1000 loops, best of 100: 603 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 y.backward(retain_graph=True)
100 loops, best of 100: 13.3 ms per loop
>>> m = nn.PReLU(100)
>>> x = torch.randn(100, 100, 100, requires_grad=True)
>>> %timeit -r 100 y = m(x)
1000 loops, best of 100: 655 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 y.backward(retain_graph=True)
100 loops, best of 100: 2.45 ms per loop
```
CUDA with weight.numel() = 1
```
>>> m = nn.PReLU().cuda()
>>> x = torch.randn(100, 100, 100, requires_grad=True).cuda()
>>> %timeit -r 100 torch.cuda.synchronize(); y = m(x); torch.cuda.synchronize();
10000 loops, best of 100: 187 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 torch.cuda.synchronize(); y.backward(retain_graph=True); torch.cuda.synchronize();
100 loops, best of 100: 2.01 ms per loop
>>> m = nn.PReLU().cuda()
>>> x = torch.randn(100, 100, 100, requires_grad=True).cuda()
>>> %timeit -r 100 torch.cuda.synchronize(); y = m(x); torch.cuda.synchronize();
1000 loops, best of 100: 195 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 torch.cuda.synchronize(); y.backward(retain_graph=True); torch.cuda.synchronize();
100 loops, best of 100: 2.28 ms per loop
```
CUDA with weight.numel() = channel
```
>>> m = nn.PReLU(100).cuda()
>>> x = torch.randn(100, 100, 100, requires_grad=True).cuda()
>>> %timeit -r 100 torch.cuda.synchronize(); y = m(x); torch.cuda.synchronize();
1000 loops, best of 100: 174 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 torch.cuda.synchronize(); y.backward(retain_graph=True); torch.cuda.synchronize();
100 loops, best of 100: 2.27 ms per loop
>>> m = nn.PReLU(100).cuda()
>>> x = torch.randn(100, 100, 100, requires_grad=True).cuda()
>>> %timeit -r 100 torch.cuda.synchronize(); y = m(x); torch.cuda.synchronize();
10000 loops, best of 100: 181 µs per loop
>>> y = m(x).sum()
>>> %timeit -r 100 torch.cuda.synchronize(); y.backward(retain_graph=True); torch.cuda.synchronize();
100 loops, best of 100: 2.26 ms per loop
```
The huge performance regression in CPU when weight.numel() = 1 is addressed by replacing at::CPU_tensor_apply* with parallelized kernels.
ezyang SsnL zou3519 soumith
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11758
Differential Revision: D9995799
Pulled By: weiyangfb
fbshipit-source-id: d289937c78075f46a54dafbde92fab0cc4b5b86e
Summary:
- fix PR https://github.com/pytorch/pytorch/pull/11061 by moving `detach_()` and `set_requires_grad()` to `torch.tensor_ctor()` and `tensor.new_tensor`, and also removed warnings and `args_requires_grad` from `internal_new_from_data `
- with this patch, the returned tensor from `tensor_ctor()` and `new_tensor` will be detached from source tensor, and set requires_grad based on the input args
- `torch.as_tensor` retains its behavior as documented
gchanan apaszke
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11815
Differential Revision: D9932713
Pulled By: weiyangfb
fbshipit-source-id: 4290cbc57bd449954faadc597c24169a7b2d8259
Summary:
+ https://github.com/pytorch/pytorch/issues/10236 : torch.bernoulli's out kwarg is broken
fixed in moving `bernoulli_out` to ATen
+ https://github.com/pytorch/pytorch/issues/9917 : BUG torch.bernoulli(p.expand(shape)) is broken
fixed in moving all `bernoulli` ops in ATen to use the modern apply utils methods
+ https://github.com/pytorch/pytorch/issues/10357 : torch.bernoulli inconsistent gpu/cpu results
fixed by adding CUDA asserts
In order to use `curand_uniform4`, I made some changes to `CUDAApplyUtils.cuh`. Specifically, I introduced an optional template parameter `int step` to the `CUDA_tensor_applyN` methods, representing that we want to process `step` values at each time for each of the `N` tensors.
The calling convention for `step = 1` (default) isn't changed. But if `step > 1`, the given lambda `op` must take in `int n` as its first argument, representing the number of valid values, because there may not be full `step` values at the boundary. E.g., here is what the `bernoulli(self, p_tensor)` call look like:
```cpp
// The template argument `4` below indicates that we want to operate on four
// element at each time. See NOTE [ CUDA_tensor_applyN helpers ] for details.
at::cuda::CUDA_tensor_apply2<scalar_t, prob_t, 4>(
ret, p,
[seeds] __device__(
int n, scalar_t& v1, scalar_t& v2, scalar_t& v3, scalar_t& v4,
const prob_t& p1, const prob_t& p2, const prob_t& p3, const prob_t& p4) {
curandStatePhilox4_32_10_t state;
curand_init(
seeds.first,
blockIdx.x * blockDim.x + threadIdx.x,
seeds.second,
&state);
float4 rand = curand_uniform4(&state);
switch (n) {
case 4: {
assert(0 <= p4 && p4 <= 1);
v4 = static_cast<scalar_t>(rand.w <= p4);
}
case 3: {
assert(0 <= p3 && p3 <= 1);
v3 = static_cast<scalar_t>(rand.z <= p3);
}
case 2: {
assert(0 <= p2 && p2 <= 1);
v2 = static_cast<scalar_t>(rand.y <= p2);
}
case 1: {
assert(0 <= p1 && p1 <= 1);
v1 = static_cast<scalar_t>(rand.x <= p1);
}
}
}
);
```
Benchmarking on `torch.rand(200, 300, 400)` 20 times, each time with 20 loops:
post patch
```
➜ ~ numactl --cpunodebind 1 --membind 1 -- taskset -c 12,13,14,15,16,17,18,19,20,21,22,23 env CUDA_LAUNCH_BLOCKING=1 python bern.py
torch.bernoulli(x)
6.841588497161865 +- 0.05413117632269859
torch.bernoulli(xc)
0.05963418632745743 +- 0.0008014909108169377
x.bernoulli_()
0.4024486541748047 +- 0.0021550932433456182
xc.bernoulli_()
0.02167394384741783 +- 2.3818030967959203e-05
```
pre-patch
```
➜ ~ numactl --cpunodebind 1 --membind 1 -- taskset -c 12,13,14,15,16,17,18,19,20,21,22,23 env CUDA_LAUNCH_BLOCKING=1 python bern.py
torch.bernoulli(x)
12.394511222839355 +- 0.0966421514749527
torch.bernoulli(xc)
0.08970972150564194 +- 0.0038722590543329716
x.bernoulli_()
1.654480218887329 +- 0.02364428900182247
xc.bernoulli_()
0.058352887630462646 +- 0.003094920190051198
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10273
Differential Revision: D9831294
Pulled By: SsnL
fbshipit-source-id: 65e0655a36b90d5278b675d35cb5327751604088
Summary:
Adds vararg support for meshgrid and adds checks for all the tensor arguments to have the same dtype and device.
Fixes: [#10823](https://github.com/pytorch/pytorch/issues/10823), #11446
The earlier pull request closed without any changes because I had some rebasing issues, so I made another pull request to close out #10823. Sorry for the inconvenience.
Differential Revision: D9892876
Pulled By: ezyang
fbshipit-source-id: 93d96cafc876102ccbad3ca2cc3d81cb4c9bf556
Summary:
tset_potri -> test_potri, even though it has been like this for a long time
More a curiosity than grave functionality...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11770
Reviewed By: ezyang
Differential Revision: D9884767
Pulled By: soumith
fbshipit-source-id: 9bedde2e94ade281ab1ecc2293ca3cb1a0107387
Summary:
Fixes#11663
`TensorIterator` was replacing the op tensors with type casted tensors
which ended up producing side effects in binary ops like `a.float() * b`
where `a` and `b` are `LongTensor`s.
colesbury ezyang apaszke
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11708
Differential Revision: D9834016
Pulled By: driazati
fbshipit-source-id: 4082eb9710b31dfc741161a0fbdb9a8eba8fe39d
Summary:
…cuda())
While I was at it, I audited all other ways I know how we might get a CUDA
type from PyTorch and fixed more constructors which don't work.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11533
Differential Revision: D9775786
Pulled By: ezyang
fbshipit-source-id: cd07cdd375fdf74945539ec475a48bf08cbc0c17
Summary:
Arg parser allowed additional positional args to be parsed into keyword-only params.
Fixes a couple cases:
- The positional argument happens to be of the right type, and it just works silently. Now, we fail as expected.
- The positional argument fails later down the line. Now, we fail at the appropriate time and get a better error message.
Pre-fix:
```
>>> torch.cuda.LongTensor((6, 0), 1, 1, 0)
tensor([6, 0], device='cuda:1')
```
Post-fix:
```
>>> torch.cuda.LongTensor((6, 0), 1, 1, 0)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: new() received an invalid combination of arguments - got (tuple, int, int, int), but expected one of:
* (torch.device device)
* (torch.Storage storage)
* (Tensor other)
* (tuple of ints size, torch.device device)
* (object data, torch.device device)
```
Pre-fix:
```
>>> a = torch.tensor(5)
>>> a.new_zeros((5,5), 0)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: new_zeros(): argument 'dtype' (position 2) must be torch.dtype, not int
```
Post-fix:
```
>>> a = torch.tensor(5)
>>> a.new_zeros((5,5), 0)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: new_zeros() takes 1 positional argument but 2 were given
```
fixes#8351
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10499
Differential Revision: D9811093
Pulled By: li-roy
fbshipit-source-id: ce946270fd11b264ff1b09765db3300879491f76
Summary:
After discussions in #11584 , new PR for just the test skip and hgemm integration.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11593
Differential Revision: D9798527
Pulled By: ezyang
fbshipit-source-id: e2ef5609676571caef2f8e6844909fe3a11d8b3e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11420
Surprisingly tricky! Here are the major pieces:
- We grow a even yet more ludicrous macro
AT_FORALL_SCALAR_TYPES_WITH_COMPLEX_EXCEPT_COMPLEX_HALF
which does what it says on the tin. This is because I was
too lazy to figure out how to define the necessary conversions
in and out of ComplexHalf without triggering ambiguity problems.
It doesn't seem to be as simple as just Half. Leave it for
when someone actually wants this.
- Scalar now can hold std::complex<double>. Internally, it is
stored as double[2] because nvcc chokes on a non-POD type
inside a union.
- overflow() checking is generalized to work with complex.
When converting *to* std::complex<T>, all we need to do is check
for overflow against T. When converting *from* complex, we
must check (1) if To is not complex, that imag() == 0
and (2) for overflow componentwise.
- convert() is generalized to work with complex<->real conversions.
Complex to real drops the imaginary component; we rely on
overflow checking to tell if this actually loses fidelity. To get
the specializations and overloads to work out, we introduce
a new Converter class that actually is specializable.
- Complex scalars convert into Python complex numbers
- This probably fixes complex tensor printing, but there is no way
to test this right now.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Reviewed By: cpuhrsch
Differential Revision: D9697878
Pulled By: ezyang
fbshipit-source-id: 181519e56bbab67ed1e5b49c691b873e124d7946
Summary:
vishwakftw Your patch needed some updates because the default native function dispatches changed from `[function, method]` to `[function]`. The CI was run before that change happened so it still shows green, but the internal test caught it.
I did some changes when rebasing and updating so I didn't just force push to your branch. Let's see if this passes CI and internal test. If it does, let me know if you want me to force push to your branch or use this PR instead.
Note to reviewers: patch was already approved at #10068 .
cc yf225
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11421
Differential Revision: D9733407
Pulled By: SsnL
fbshipit-source-id: cf2ed293bb9942dcc5158934ff4def2f63252599
Summary:
This PR cleans up the `at::Tensor` class by removing all methods that start with an underscore in favor of functions in the `at::` namespace. This greatly cleans up the `Tensor` class and makes it clearer what is the public and non-public API.
For this I changed `native_functions.yaml` and `Declarations.cwrap` to make all underscore methods `variant: function` (or add such a statement to begin with), and then fixed all code locations using the underscore methods.
ezyang colesbury gchanan
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11152
Differential Revision: D9683607
Pulled By: goldsborough
fbshipit-source-id: 97f869f788fa56639c05a439e2a33be49f10f543
Summary:
Add the gpu kernel version.
The parallelism I went with performs poorly when there are a large number of vectors, but they're all short, as I don't allocate the thread pool to wrap in that case.
Test Plan
---------
```
python -m unittest test_torch.TestTorch.test_pdist_{empty,scipy} test_nn.TestNN.test_pdist{,_zeros,_empty_row,_empty_col,_cpu_gradgrad_unimplemented,_cuda_gradgrad_unimplemented} test_jit.TestJitGenerated.test_nn_pdist
```
Current performance specs are a little underwhelming, I'm in the process of debugging.
size | torch | torch cuda | scipy
-----|-------|------------|------
16 x 16 | 9.13 µs ± 3.55 µs | 9.86 µs ± 81.5 ns | 15.8 µs ± 1.2 µs
16 x 1024 | 15 µs ± 224 ns | 9.48 µs ± 88.7 ns | 88.7 µs ± 8.83 µs
1024 x 16 | 852 µs ± 6.03 µs | 7.84 ms ± 6.22 µs | 4.7 ms ± 166 µs
1024 x 1024 | 34.1 ms ± 803 µs | 11.5 ms ± 6.24 µs | 273 ms ± 6.7 ms
2048 x 2048 | 261 ms ± 3.5 ms | 77.5 ms ± 41.5 µs | 2.5 s ± 97.6 ms
4096 x 4096 | 2.37 s ± 154 ms | 636 ms ± 2.97 µs | 25.9 s ± 394 ms
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11102
Differential Revision: D9697305
Pulled By: erikbrinkman
fbshipit-source-id: 2b4f4b816c02b3715a85d8db3f4e77479d19bb99
Summary:
In #9466 I got rid of storage views and eliminated all places where
they were used... OR SO I THOUGHT. In actuality, under certain
conditions (specifically, if you trained a CUDA multiprocessing model
shared over CUDA IPC and then serialized your parameters), you could
also serialize storage slices to the saved model format. In #9466,
I "fixed" the case when you loaded the legacy model format (really,
just unshared the storages--not strictly kosher but if you aren't
updating the parameters, shouldn't matter), but NOT the modern model format, so
such models would fail.
So, I could have applied the legacy model format fix too, but
hyperfraise remarked that he had applied a fix that was effectively
the same as unsharing the storages, but it had caused his model to
behave differently. So I looked into it again, and realized that
using a custom deleter, I could simulate the same behavior as old
storage slices. So back they come.
In principle, I could also reimplement storage views entirely using
our allocators, but I'm not going to do that unless someone really
really wants it.
Fixes#10120.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11314
Reviewed By: ailzhang
Differential Revision: D9671966
Pulled By: ezyang
fbshipit-source-id: fd863783d03b6a6421d6b9ae21ce2f0e44a0dcce
Summary:
Allows mulitplication of e.g. numpy.float32 with tensors.
This came up with #9468
If you want this and after the other patch is done, I'll add tests (but that would be conflicting, so I prefer to wait).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9659
Differential Revision: D8948078
Pulled By: weiyangfb
fbshipit-source-id: c7dcc57b63e2f100df837f70e1299395692f1a1b
Summary:
* improve docker packages (install OpenBLAS to have at-compile-time LAPACK functionality w/ optimizations for both Intel and AMD CPUs)
* integrate rocFFT (i.e., enable Fourier functionality)
* fix bugs in ROCm caused by wrong warp size
* enable more test sets, skip the tests that don't work on ROCm yet
* don't disable asserts any longer in hipification
* small improvements
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10893
Differential Revision: D9615053
Pulled By: ezyang
fbshipit-source-id: 864b4d27bf089421f7dfd8065e5017f9ea2f7b3b
Summary:
Also add single grad whitelist to the jit test
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10782
Reviewed By: ezyang
Differential Revision: D9583378
Pulled By: erikbrinkman
fbshipit-source-id: 069e5ae68ea7f3524dec39cf1d5fe9cd53941944
Summary:
Initial version of `unique` supporting a `dim` argument.
As discussed in [this issue](https://github.com/pytorch/pytorch/issues/9997) I added the `dim` argument to `torch.unique` with the same behavior like [numpy](https://docs.scipy.org/doc/numpy-1.14.0/reference/generated/numpy.unique.html).
Since the implementation is based on `std/thrust::unique`, the `tensor` always needs to be sorted. The `sorted` argument in `torch.unique` does not have any function, as in the CUDA version of the plain `torch.unique`.
To check the performance and equal behavior between `torch.unique` and `np.unique`, I've used [this gist](https://gist.github.com/ptrblck/ac0dc862f4e1766f0e1036c252cdb105).
Currently we achieve the following timings for an input of `x = torch.randint(2, (1000, 1000))`:
(The values are calculated by taking the average of the times for both dimension)
| Device | PyTorch (return_inverse=False) | Numpy (return_inverse=False) | PyTorch (return_inverse=True) | Numpy (return_inverse=True) |
| --- | --- | --- | --- | --- |
| CPU | ~0.007331s | ~0.022452s | ~0.011139s | ~0.044800s |
| GPU | ~0.006154s | - | ~0.105373s | - |
Many thanks to colesbury for the awesome mentoring and the valuable advices on the general implementation and performance issues!
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10423
Differential Revision: D9517289
Pulled By: soumith
fbshipit-source-id: a4754f805223589c2847c98b8e4e39d8c3ddb7b5
Summary:
Fix#10345, which only happens in CUDA case.
* Instead of returning some random buffer, we fill it with zeros.
* update torch.symeig doc.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10645
Reviewed By: soumith
Differential Revision: D9395762
Pulled By: ailzhang
fbshipit-source-id: 0f3ed9bb6a919a9c1a4b8eb45188f65a68bfa9ba
Summary:
When 0-sized dimension support is added, we expect an empty sparse tensor to be a 1-dimensional tensor of size `[0]`, with `sparseDims == 1` and `denseDims == 0`. Also, we expect the following invariants to be preserved at all times:
```
_sparseDims + _denseDims = len(shape)
_indices.shape: dimensionality: 2, shape: (_sparseDims, nnz)
_values.shape: dimensionality: 1 + _denseDims. shape: (nnz, shape[_sparseDims:])
```
This PR fixes various places where the invariants are not strictly enforced when 0-sized dimension support is enabled.
Tested and `test_sparse.py` passes locally on both CPU and CUDA with the `USE_TH_SIZE_ZERO_DIM` flag.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9279
Differential Revision: D8936683
Pulled By: yf225
fbshipit-source-id: 12f5cd7f52233d3b26af6edc20b4cdee045bcb5e
Summary:
Implemented via a wrapper, thank you Richard for the suggestion!
Fixes: #9929
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10067
Differential Revision: D9083388
Pulled By: soumith
fbshipit-source-id: 9ab21cd35278b01962e11d3e70781829bf4a36da
Summary:
Test only for existence for now. I had to skip a lot of them so there a FIXME in the test.
Also I'm not testing torch.* because of namespace issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10311
Differential Revision: D9196341
Pulled By: SsnL
fbshipit-source-id: 9c2ca1ffe660bc1cc664474993f8a21198525ccc
Summary:
* some small leftovers from the last PR review
* enable more unit test sets for CI
* replace use of hcRNG w/ rocRAND (docker image was already updated w/ newer rocRAND)
* use rocBLAS instead of hipBLAS to allow convergence w/ Caffe2
* use strided_batched gemm interface also from the batched internal interface
* re-enable Dropout.cu as we now have philox w/ rocRAND
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10406
Reviewed By: Jorghi12
Differential Revision: D9277093
Pulled By: ezyang
fbshipit-source-id: 7ef2f6fe4ead77e501ed7aea5c3743afe2466ca2
Summary:
This PR for the ROCm target does the following:
* enable some unit tests on ROCm
* fix a missing static_cast that breaks BatchNorm call on ROCm
* fix BatchNorm to work on ROCm w/ ROCm warp sizes etc
* improve the pyhipify script by introducing kernel scope to some transpilations and other improvements
* fix a linking issue on ROCm
* for more unit test sets: mark currently broken tests broken (to be fixed)
* enable THINLTO (phase one) to parallelize linking
* address the first failing of the elementwise kernel by removing non-working ROCm specialization
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10266
Differential Revision: D9184178
Pulled By: ezyang
fbshipit-source-id: 03bcd1fe4ca4dd3241f09634dbd42b6a4c350297
Summary:
This exposes expand_outplace to python. Fixes#8076. Fixes#10041.
I didn't name it torch.broadcast because numpy.broadcast does something
slightly different (it returns an object with the correct shape
information).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10075
Differential Revision: D9125816
Pulled By: zou3519
fbshipit-source-id: ebe17c8bb54a73ec84b8f76ce14aff3e9c56f4d1
Summary:
This causes numpy to yield to the torch functions,
e.g. instead of numpy array/scalar __mul__ converting the tensor to
an array, it will now arrange for the Tensor __rmul__ to be called.
Fixes case 2 of #9468
I also makes case 3 and 4 equivalent but does not fix them.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9651
Differential Revision: D8948079
Pulled By: ezyang
fbshipit-source-id: bd42c04e96783da0bd340f37f4ac3559e9bbf8db
Summary:
These could use some autograd tests, which are coming in a later PR, but using them in autograd is probably pretty rare.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9947
Reviewed By: ezyang
Differential Revision: D9032778
Pulled By: gchanan
fbshipit-source-id: fa5a6509d3bac31ea4fae25143e82de62daabfbd
Summary:
```
This adds TensorIterator, a helper class for computing element-wise
operations that's intended to replace the CPU and CUDA apply utils
functions.
CPU kernels are implemented as functions that operate on strided 1-d
tensors compared to CPUApplyUtils which operated individual elements. This
allows the kernels to handle vectorization, while TensorIterator handles
parallelization and non-coalesced dimensions.
GPU kernels continue to operate on elements, but the number of
specializations is reduced. The contiguous case remains the same. The
non-contiguous case uses a single (reduced) shape for all operands and
the fast integer division from THCIntegerDivider. To avoid extra
specializations for indexing with 64-bits, large operations are split
into smaller operations that can be indexed with 32-bits.
Major semantic changes:
- No more s_add, s_mul, s_div, or s_sub. Broadcasting is handled by
TensorIterator. The autograd engine performs the reduction assuming
standard broadcasting if the gradient shape does not match the
expected shape. Functions that do not use standard broadcasting rules
should either continue to trace the expand calls or handle the
reduction in their derivative formula.
- Use ONNX v7, which supports broadcasting ops.
Performance impact:
- Small increased fixed overhead (~0.5 us)
- Larger overhead for wrapped numbers (~2.5 us)
- No significant change for ops on contiguous tensors
- Much faster worst-case performance for non-contiguous GPU tensors
- Faster CPU bias addition (~2x)
- Faster GPU bias addition (~30% faster)
Future work:
- Decrease overhead, especially for wrapping numbers in Tensors
- Handle general inter-type operations
- Extend to unary ops and reductions
- Use buffering for compute-bound operations on non-contiguous tensors
(pull in from CPUApplyUtils)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/8919
Differential Revision: D8677600
Pulled By: colesbury
fbshipit-source-id: 61bc9cc2a36931dfd00eb7153501003fe0584afd
Summary:
The primary use-site of typeString was checked_cast_tensor.
I did a little more than I needed in this patch, to set
the stage for actually deleting the tensor type.
Specifically, I modified checked_cast_tensor to explicitly
take Backend and ScalarType, the idea being that once we
remove the tensor subclasses, we will delete the T template
parameter.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9764
Differential Revision: D8969196
Pulled By: ezyang
fbshipit-source-id: 9de92b974b2c28f12ddad13429917515810f24c6
Summary:
Fixes: #9754
Maybe this could also make its way into 0.4.1, it is a severe debugging headache if you hit this...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9755
Reviewed By: ezyang
Differential Revision: D8967178
Pulled By: zou3519
fbshipit-source-id: 151ed24e3a15a0c67014e411ac808fb893929a42
Summary:
…unctions.
This also unifies the error checkign between scatter/scatterAdd on CUDA.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9658
Differential Revision: D8941527
Pulled By: gchanan
fbshipit-source-id: 750bbac568f607985088211887c4167b67be11ea
Summary:
This is mainly straightforward, with two exceptions:
1) cublasSgemv, cublasDgemv appear to have a bug where (x,0).mv(0) does not handle beta, whereas cublasSgemm, cublasDgemm do for case where (x,0).mm(0,y). This is handled by manually calling zero / mul.
2) I fixed a bug in btrifact that was broken even when dealing with non-empty tensors. Basically, if out.stride(0) was 1, because the underlying BLAS call expects column-major matrices, to get a column-major tensor, out.transpose_(0, 1) would be called. But this is just wrong, as if the batch dimension (0) doesn't match the size of the columns (1), you don't even have a tensor of the correct shape.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9573
Reviewed By: ezyang
Differential Revision: D8906144
Pulled By: gchanan
fbshipit-source-id: de44d239a58afdd74d874db02f2022850dea9a56
Summary:
…CPU LAPACK routines.
Note that the LAPACK functions in general require a different approach, because direct calls with size zero dims do not work.
Here I just selected a reasonable subset of LAPACK routines to support.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9522
Reviewed By: ezyang
Differential Revision: D8888180
Pulled By: gchanan
fbshipit-source-id: 16b9013937806d375d83d1c406815765fda00602
Summary:
If this is good, I could write some tests to ensure collision doesn't occur within a given range.
Closes#7228
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9246
Differential Revision: D8872608
Pulled By: ezyang
fbshipit-source-id: 0ed29a73188f4167b42756f59a5c9a3d5cb37326
Summary:
…ors (CPU).
This includes (mainly) CPU fixes; CUDA fixes are a little more involved because you can't use an empty grid.
This also includes a fix for index_copy, which checked that self.size(dim) == src.size(0), which isn't correct (the same dimension should be compared).
Finally, also includes a fix for CUDA flip (although it's not tested yet), to get the stride using multiplication rather than division to avoid divide-by-0.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9482
Reviewed By: ezyang
Differential Revision: D8873047
Pulled By: gchanan
fbshipit-source-id: 86523afd3d50277834f654cd559dfbc7875cdffe
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9497Fixes#7883 by using `rfft`.
It's worth noting that this is BC breaking. And it's impossible to detect the change because the two signatures before and after this change supports a common subset of calling patterns, e.g., `stft(Tensor, int, int)`. (some other calling patterns will raise error).
soumith and I plan to change the current `stft` interface because it is a bit messy and non-standard. rafaelvalle suggested us that `librosa` is a good reference API to align with. After discussing with soumith and ezyang , and given that `stft` is only out for 1 release, I decide to go with directly changing the signature. Also, my understanding is that most researchers in this field will welcome this change as `librosa` seems to be the golden-standard here. (it doesn't yet support all `pad_mode` but those will become available if added to `F.pad`.)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9308
Reviewed By: ezyang
Differential Revision: D8806148
Pulled By: SsnL
fbshipit-source-id: f6e8777d0c34d4a4d7024e638dc9c63242e8bb58
Summary:
Storage views were previously used to implement CUDA IPC sharing,
but they weren't necessary. The new strategy is described in
Note [CUDA IPC and the caching allocator].
This also fixes an unrelated bug, where we weren't actually using
the Tensor forking pickler, because we didn't register a pickler
for torch.Tensor.
Fixes#9447. Fixes#46.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
CC apaszke
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9466
Reviewed By: apaszke
Differential Revision: D8859698
Pulled By: ezyang
fbshipit-source-id: 3362cb92f6ae4aa37084c57d79b31004bd0b4a97
Summary:
`test_neg` sometimes fails internally because `random_()` can generate an out-of-range value for CharTensor. This PR fixes it.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9431
Reviewed By: SsnL
Differential Revision: D8843284
Pulled By: yf225
fbshipit-source-id: bf516cceb8f780e133fa54f7364c77821eb7c013
Summary:
Fixes: #9421
I don't think it is easy to deal with non-contiguous array in cuda topk, so I'm adding a check.
The argument number is a bit confusing when it shows in PyTorch but it is consistent with the other checks. (Not sure whether it would make sense to eliminate argument numbers from the error TH/THC error messages given that they're probably off more than once...)
Do we need a test that it indeed refuses non-contiguous?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9441
Reviewed By: soumith
Differential Revision: D8850719
Pulled By: ezyang
fbshipit-source-id: d50561bb37ed50ab97aeaf54d8e3fc6c765bdc7c
Summary:
This includes either bug fixes or NumPy semantics changes for the following methods:
chunk, diagonal, unfold, repeat, flatten, reshape, split, unsqueeze.
The n-dimensional empty tensor feature is still hidden behind a feature flag.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9362
Reviewed By: ezyang
Differential Revision: D8817002
Pulled By: gchanan
fbshipit-source-id: 6ff704ec96375f00b4dd39ebcd976efac0607fb4
Summary:
This PR implements and tests N-dimensional empty tensors for indexing, factories, and reductions if compiled with -DUSE_TH_SIZE_ZERO_DIM.
Still remaining to add:
1) TensorShape functions
2) Simple linear algebra functions (matrix multiply variants)
3) Other functions that operate over a dimension (but don't reduce).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/9209
Reviewed By: ezyang
Differential Revision: D8751257
Pulled By: gchanan
fbshipit-source-id: 2113374dc7af6caf31a99bf67b3893f130a29e23
Summary:
cc vishwakftw
Also added a check if none of the input tensors in `gradcheck` have `requires_grad=True`.
Closes https://github.com/pytorch/pytorch/pull/9192
Differential Revision: D8739401
Pulled By: SsnL
fbshipit-source-id: 81bb3aa0b5c04eb209b137a4bd978e040e76cbcd
Summary:
Closes#9147
Added a test to prevent regression in test_torch
Added entries in docs
cc ezyang weiyangfb
Closes https://github.com/pytorch/pytorch/pull/9156
Differential Revision: D8732095
Pulled By: soumith
fbshipit-source-id: 7a6892853cfc0ccb0142b4fd25015818849adf61
Summary:
This will resolve some of the timeout issues in CPU and GPU tests internally.
Closes https://github.com/pytorch/pytorch/pull/9061
Reviewed By: ezyang
Differential Revision: D8707471
Pulled By: yf225
fbshipit-source-id: 9dc82a2c9da0c540ae015442f74b9b2b1a67a246
Summary:
Currently the `test_RNG_after_pickle` in the PR would fail because pickling a tensor changes the RNG state. This PR aims to fix it.
Closes https://github.com/pytorch/pytorch/pull/8971
Reviewed By: ezyang
Differential Revision: D8677474
Pulled By: yf225
fbshipit-source-id: 1713d9611699ad288b66d92dbb29ce9feb34b8cf
* add opencl + fpga context
adds an opencl context inside caffe2/fb which can be used for fpga access
* [Caffe2] Force tensor inference checks to be triggered during testing
We've started to rely on TensorInference functions more for different analysis. This diff ensures that the TensorInference function's result matches what is expected from the definition of the operator.
* Enable building //caffe2:torch with @mode/opt
In @mode/opt, python runs out of a PAR, which breaks a lot of
assumptions in the code about where templates/ folders live relative
to __file__. Rather than introduce hacks with parutil, I simply turn
template_path into a parameter for all the relevant functions and
thread it through from the top level.
* [Caffe2] Fix cost models for DotProduct and Div. Update Tensor Inference for dot product
As title. DotProduct states that output is a 1-D tensor (https://caffe2.ai/docs/operators-catalogue.html#dotproduct) though code suggests it is either 0- or 1-D depending on inputs. TensorInference defined to support implementation.
* [SG-MoE] Add an option to make the experts NOT as components
* [nomnigraph] Rename and fixup convertToNeuralNetOperator API
This will make things a bit cleaner
* no longer symlink THNN.h and THCUNN.h
* forced decoder network (onnx export)
Closes https://github.com/pytorch/translate/pull/95
Add networks in ensemble_export.py to create a forced decoding network from PyTorch NMT checkpoints. This network takes an arbitrary numberized (source, target) pair and returns the model score for the translation, including penalties.
Vocabulary reduction networks are also supported, but note that target indices which are not in the possible_translation_tokens generated for the source input will be trea
* Revert schema change to fix production models
Revert schema change to fix production models
* MockLogDeviceReader - rebase on FIX
# Goal
1), Build a make_mock_log_device_reader using make_mock_reader
2), Replace the real log_device_reader here: https://fburl.com/raihwf1p
# Log by D8151734
Real log_device_reader:
```
I0529 20:29:05.373108 954994 tensor.h:839] Tensor print_net/log of type std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >. Dims: (): read_net/ParseOpenTrainingRow:0
I0529 20:29:05.373244 954994 tensor.h:839] Tensor read_net/ParseOpenTrainin
* [C2/D2][1/n]: Nonnegative-Constrained Optimization -- log barrier
implement log barrier as a regularization method
* Add teacher weight screening.
Add teacher weight sceening according to teacher labels. If teacher label is zero, we do not use the distill loss in the objective function.
* Add NormalizerContext
See task for more detail. This implementation is a copy of what exists for RegularizerContext except for how the parameters are defined in the model_definition thrift file.
I'll try an alternative implementation which overrides the default arguments of functions instead like for argscopes in tensorflow.
https://github.com/pytorch/pytorch/compare/master...MaximeBoucher:update-from-facebook-0939578c068c?expand=1
* Adding cosine similarity option in dot processor
Add pairwise cosine similarity option in dot product.
Add an option to concate dot product and cosine similarity.
Add test cases.
* [nomnigraph][redo] Concat elim for sparseNN
Same as D7962948, which was reverted because Operator Schema was not
defined
* [pytorch] Revert pytorch/pytorch#7918 'Release GIL when copying to shared memory', breaks ASAN
Revert this pytorch diff that breaks ASAN when running Filament in dev mode; in opt mode it gives "bad file descriptor" errors. Looks like a race when copying tensors to shared memory in multiple mp.Queue's (which spawn separate threads).
https://github.com/pytorch/pytorch/pull/7918/files
* [nomnigraph][mobile] Enable nomnigraph by default, use -Oz on nomnigraph related code to reduce code size
enables nomnigraph and reduces codesize
* [Warmup] Allow both offline incremental training and online training
Change plan name on saving side and reading side to support both training type
This diff depends on D8128530 and D8168651.
* Revert D7802642: [Warmup] Allow both offline incremental training and online training
This reverts commit afc213cf9b36cecf75333a788391c4d09f4afccc
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* Add legacy grad logic to fix div op on old graphs.
Add legacy grad logic to fix div op on old graphs.
* Correctly propagate operator failures
Propagate errors from operators that throw exceptions and return false
* Revert D8374829: [caffe2][nomnigraph][redo] Concat elim for sparseNN
This reverts commit 6dda028c463e54bb5c32188bbbe9202107e188a5
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* [Caffe2] Added extra_info to core.DeviceOption(), enforced extra_info to be inherited in scope.DeviceScope
extra_info is a newly defined field in DeviceOption proto. This diff added extra_info to the core.DeviceOption(). And, In scope.DeviceScope(), this diff enforce the new scope to inherit the extra_info from old scope.
* [opt] hgdirsync wasn't enabled, merge diverged code
Here's the damage, P59732616 basically xplat was left behind but had
the change from assert to CAFFE_ENFORCE
* OMP parallelism over RoIs for RoIAlign op
Simpler to parallelize over RoIs. Shouldn't affect other uses as it relies on
the number of OMP threads set during startup.
PR: https://github.com/pytorch/pytorch/pull/8562
* Use int64_t for shape in FillOps
to avoid overflow of int32
* Implement Rotated RoIAlign op
Based on Rotated RPNs as explained in https://arxiv.org/abs/1703.01086.
The idea is simple - orientation/angle is added as an RPN
anchor parameter and then the angle is further regressed similar to bbox
coords. There are some additional changes related to NMS and IoU, but besides
that it's a direct extension to Faster-RCNN. Further details in https://fb.quip.com/sZHlA1iMfWPZ.
RoIs are represented in [center_x, center_y, width, height, angle] format.
`angle` repre
* Rotated RoIAlign op CUDA forward implementation
CUDA forward impl for D8415490
* RoIAlignRotated op CUDA backward pass implementation
TSIA
* All remaining fixes to eliminate process_github.sh
Most of this diff has already been reviewed separately, except for the parts relating to _thnn/utils.py and _utils._internal.py
remove skipIf(True, 'Fbcode') line from process_github.sh
replace sed of cpp file with #ifdef to control cudnnDestroy use
undo sync-time deletion of .gitattributes, remove process_github.sh
switch to using _utils._internal rather than try-import-except
This diff also fixes the open-source bug where rebuilds have
* Back out "Revert D7802642: [Warmup] Allow both offline incremental training and online training"
Original commit changeset: 7707d2efe60e The original diff is backout becuase the online trainer package is backed out. This code would only work with new online trainer package
* [easy] improve error log in adagrad op
as title
* re-allow use of thnn_h_path
This fixes cffi usage in OSS
* [4/4] [tum] paralyzing layerNorm for GPU full sync
as title
* add compile=False to pytorch tests, remove hack with pyc
* Add shape and type inference for RowWiseArgMax operator
See title
* Revert D8515341: Back out "Revert D7802642: [Warmup] Allow both offline incremental training and online training"
This reverts commit 78167eeef0af16b60f72c82f9dcdda9b41b4dcbd
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* [fix-flaky-test] mock_hive_reader_test flaky, because GlobalCounter collects local counts intervally
# Problem
`MockHiveReader` uses `GlobalCounter` to limit `max_examples`.
GlobalCounter on server node collect local counts from worker nodes every 1 sec.
This 1 sec delay makes it impossible to limit exactly to the `max_examples`, it will definitely exceed `max_examples`.
# Plan
Given,
```
Expected num_examples = max_examples + num_examples/sec (Read Speed) x 1 sec (GlobalCounter Sync Int
* [Caffe2] Fix FCGradient cost inference. Prevent overflow in cost inference
FCGradient missed a factor 2 in the `num_outputs == 3` case. Overflow was occurring with flop calculation for FC. Changed types to `uint64_t` to prevent future problems.
* Fix binary ops with empty inputs
Fix binary ops with empty inputs
* Support the filling of input blob with provided data
as title for Biz Integrity case
* Back out "Revert D8515341: Back out "Revert D7802642: [Warmup] Allow both offline incremental training and online training""
Original commit changeset: 30c55dd38816 Original diff is reverted due to introducing bad integration test. Fixed the integration test.
* [c2][easy] improve pack ops error loggings
as desc.
* Add ShapeTypeInference for LpNorm operator
As desc
* Shard test_nn to reduce runtime for each test target
Closes https://github.com/pytorch/pytorch/pull/8793
The current test_nn would time out and be disabled in GreenWarden, and we need to have an option to split it up in order to pass the stress test. Right now GreenWarden roughly allows running 100 test cases in test_nn before timing out, and here we have an option to divide test_nn into 30 shards (with ~40 tests in each shard) to allow for some test suite growth in the future.
* Change default caffe2_streams_per_gpu to 1
* Remove IN_SANDCASTLE from common.py and test_nn.py
We prefer to disable the failing tests through Sandcastle UI instead.
* Add a new class for an updated prof_dag.proto
This diff contains:
- An updated prof_dag.proto that contains blob profiles.
- A class to deserialize this information (serialization is in a follow up diff)
- Update to separate profiling information from NeuralNet (and use it as part of the class above).
- Unit tests
* Lambdarank for SparseNN
This diff adds a lambda_rank_layer for SparseNN.
changes include
1) Adds support for multi sessions in c2 op
2) Adds support for two different loss functions in c2 op
3) Unit tests for op
* Revert D8586950: Back out "Revert D8515341: Back out "Revert D7802642: [Warmup] Allow both offline incremental training and online training""
This reverts commit 012220ed63eccc35659a57b31d16a3625da6317b
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* [easy] A few fixups to multithread predictor benchmark
(1) support perf on T6 server
(2) remove dead code
* fix a bug about the map size
as title
* Fix reduce sum on in-place case.
Fix reduce sum on in-place case.
* [Warmup] Reland reverted diff Allow both offline incremental training and online training
Closes https://github.com/pytorch/pytorch/pull/8827
fix net transform integration test. Allow offline and online trainer to coexist D7802642.
* Add StoreHandlerNotAvailableException
Add an exception for a store that is not available or has been
deleted.
* Use exception handling for fault tolerance, missing KV store
Remove status blobs to communication ops so that exceptions propagate on
failure.
* [C2/D2][2/n]: Nonnegative-Constrained Optimization -- bounded grad proj
for simple bounded constrained optimization, incl non-negative box constraints.
* [GanH]: Adaptive Weighting with More Estimations
With implemented postivity optimization, we now learn adaptive weights with different
parameterizations.
This improves parameter estimation and training stability.
* Revert some changes for landing
* Remove AutoNoGIL in StorageSharing
* Temporarily disable net_tests
* Revert "[Caffe2] Force tensor inference checks to be triggered during testing"
This reverts commit 67ef05c22b2f71b4a489695384932f968384a2a4.
* Revert "Fix reduce sum on in-place case."
This reverts commit 6cb8a8e1b3db7b6d20941b0053e3f3836068eb64.
* Revert "Revert "Fix reduce sum on in-place case.""
This reverts commit 130a257c0893dc09f4bd6e6a45d112261807fd2c.
* Better forward methods in C++ API
capitalize error message in test_torch.test_flatten
Support for operator()
* Add operator() to Functional
* Get rid of SigmoidLinear
* Add BoundFunction to FunctionalImpl
* Remove macro from conv because it makes errors more nasty
* Some 0-sized dimension support, port catArray away from resizeLegacy.
The goal of this PR is to port catArray away from resizeLegacy (so we can delete the legacy resize calls), but since catArray has some weird behavior because
we don't have arbitrary 0-sized dimension support, I made some effort to fix these both in one pass.
The major changes here are:
1) catArray uses the new resize API, no longer the old resizeLegacy API.
2) As 1) is the last usage of resizeLegacy, it is deleted.
3) If compiled with USE_TH_SIZE_ZERO_DIM, catArray will work and properly check shapes for n-dimensional empty tensors.
4) However, we retain the old behavior of "ignoring" size [0] tensors in catArray. We previously allowed this because we didn't have n-dimensional empty tensors.
5) To get the above to work, we also add support for n-dimensional empty tensors for narrow and slice (ifdef USE_TH_SIZE_ZERO_DIM).
6) We change the stride formula for empty tensors to match NumPy; basically, we never multiply by 0 as the size, always at least 1, so the
strides are monotonically increasing in the empty tensor case.
7) We print the size of empty tensors if size != [0]; this matches NumPy behavior (even in cases where the size could be inferred from the brackets.
8) For test purposes, we add torch._C._use_zero_size_dim() to add tests for the above.
* Fix flake8.
* Address review comments.
* Created TensorOptions
Storing the type in TensorOptions to solve the Variable problem
Created convenience creation functions for TensorOptions and added tests
Converted zeros to TensorOptions
Converted rand to TensorOptions
Fix codegen for TensorOptions and multiple arguments
Put TensorOptions convenience functions into torch namespace too
All factory functions except *_like support TensorOptions
Integrated with recent JIT changes
Support *_like functions
Fix in place modification
Some cleanups and fixes
Support sparse_coo_tensor
Fix bug in Type.cpp
Fix .empty calls in C++ API
Fix bug in Type.cpp
Trying to fix device placement
Make AutoGPU CPU compatible
Remove some auto_gpu.h uses
Fixing some headers
Fix some remaining CUDA/AutoGPU issues
Fix some AutoGPU uses
Fixes to dispatch_tensor_conversion
Reset version of new variables to zero
Implemented parsing device strings
Random fixes to tests
Self review cleanups
flake8
Undo changes to variable.{h,cpp} because they fail on gcc7.2
Add [cuda] tag to tensor_options_cuda.cpp
Move AutoGPU::set_index_from into .cpp file because Windows is stupid and sucks
Fix linker error in AutoGPU.cpp
Fix bad merge conflict in native_functions.yaml
Fixed caffe2/contrib/aten
Fix new window functions added to TensorFactories.cpp
* Removed torch::TensorOptions
Added code to generate wrapper functions for factory methods
Add implicit constructor from Backend to TensorOptions
Remove Var() from C++ API and use torch:: functions
Use torch:: functions more subtly in C++ API
Make AutoGPU::set_device more exception safe
Check status directly in DynamicCUDAHooksInterface
Rename AutoGPU to DeviceGuard
Removed set_requires_grad from python_variables.h and warn appropriately in Variable::set_requires_grad
remove python_default_init: self.type()
Add back original factory functions, but with deprecation warnings
Disable DeviceGuard for a couple functions in ATen
Remove print statement
Fix DeviceGuard construction from undefined tensor
Fixing CUDA device compiler issues
Moved as many methods as possible into header files
Dont generate python functions for deprecated factories
Remove merge conflict artefact
Fix tensor_options_cuda.cpp
Fix set_requires_grad not being checked
Fix tensor_new.h
TEMPORARILY put some methods in .cpp files to see if it solves issues on windows and mac
Fix bug in DeviceGuard.h
Missing includes
TEMPORARILY moving a few more methods into .cpp to see if it fixes windows
Fixing linker errors
* Fix up SummaryOps to use new factories
Undo device agnostic behavior of DeviceGuard
Use -1 instead of optional for default device index
Also move DeviceGuard methods into header
Fixes around device index after optional -> int32_t switch
Fix use of DeviceGuard in new_with_tensor_copy
Fix tensor_options.cpp
* Fix Type::copy(
* Remove test_non_float_params from ONNX tests
* Set requires_grad=False in ONNX tests that use ints
* Put layout/dtype/device on Tensor
* Post merge fixes
* Change behavior of DeviceGuard to match AutoGPU
* Fix C++ API integration tests
* Fix flip functions
