Sam Gross
4bce69be22
Implement Variable.storage() ( #3765 )
...
This still uses THPStorage, but avoids touching THPTensor
2017-11-20 14:18:07 -05:00
Soumith Chintala
50009144c0
add warnings if device capability is less than ideal ( #3601 )
2017-11-09 11:48:59 -05:00
peterjc123
aa911939a3
Improve Windows Compatibility (for csrc/scripts) ( #2941 )
2017-11-08 19:51:35 +01:00
SsnL
bb1b826cdc
Exposing emptyCache from allocator ( #3518 )
...
* Add empty_cache binding
* cuda.empty_cache document
* update docs
2017-11-07 17:00:38 -05:00
Adam Paszke
cc3058bdac
Fix macOS build (with CUDA) ( #3071 )
2017-10-11 19:04:15 +02:00
Soumith Chintala
e9dccb3156
implement all_reduce, broadcast, all_gather, reduce_scatter
2017-10-09 22:24:18 -04:00
Soumith Chintala
4d62933529
add initial NCCL C bindings
2017-10-09 22:24:18 -04:00
Soumith Chintala
b3bc5fe302
refactor THCP method defs into cuda/Module.cpp
2017-09-30 13:14:35 -07:00
Justin Johnson
94b5990201
Add torch.cuda.get_device_name function ( #2540 )
2017-08-26 15:06:37 -04:00
Adam Paszke
8ab3d214d5
Fixes for DistributedDataParallel ( #2168 )
2017-07-21 16:00:46 -04:00
Edward Z. Yang
72e9e7abf7
Warning squash.
...
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
2017-07-21 14:13:11 -04:00
Zach DeVito
9d8cff9bc1
initialize aten and pytorch to share the same THCState
2017-07-11 10:35:03 -04:00
Adam Paszke
12813b88f6
Add DistributedDataParallel
2017-06-12 22:00:22 -04:00
Trevor Killeen
05bc877a05
make THPPointer have explicit constructors ( #1636 )
2017-05-25 15:35:54 -04:00
Sam Gross
4c1cdb6148
Refactor Python string utility function
2017-04-28 21:25:26 +02:00
Sam Gross
aab30d4ea2
Fix errors when no CUDA devices are available ( #1334 )
...
Fixes #1267
This fixes a number of issues when PyTorch was compiled with CUDA
support but run on a machine without any GPUs. Now, we treat all errors
from cudaGetDeviceCount() as if the machine has no devices.
2017-04-23 14:45:27 +02:00
Sergey Zagoruyko
8dc5d2a22e
export current_blas_handle
2017-03-23 23:32:45 +01:00
Sam Gross
b9379cfab7
Use cuDNN and NCCL symbols from _C library ( #1017 )
...
This ensures that we use the same library at the C++ level and with
Python ctypes. It moves the searching for the correct library from
run-time to compile-time.
2017-03-16 16:10:17 -04:00
soumith
7ad948ffa9
fix tests to not sys.exit(), also fix fatal error on THC initialization
2017-03-01 17:37:04 -05:00
Sam Gross
fc6fcf23f7
Lock the cudaFree mutex. ( #880 )
...
Prevents NCCL calls from overlapping with cudaFree() which can lead to
deadlocks.
2017-03-01 11:29:25 -05:00
Adam Paszke
19a65d2bea
Expose stateless methods for torch.cuda.HalfTensor
2017-02-26 20:02:42 +01:00
Sam Gross
bd5303010d
Refactor autograd package to separate Python dependencies. ( #662 )
...
The core autograd Variable, Function, and Engine no longer depend on the
Python API. This let's us implement functions in C++. In the future, we
can also multithread engine and release the GIL for most of the
non-Python backwards.
2017-02-13 16:00:16 -08:00
Zeming Lin
59d66e6963
Sparse Library ( #333 )
2017-01-05 00:43:41 +01:00
Sam Gross
20fffc8bb7
Fix torch.is_tensor for half tensors ( #322 )
...
Fixes #311
2016-12-19 15:27:47 +01:00
Sam Gross
1af9a9637f
Refactor copy and release GIL during copy ( #286 )
2016-12-11 21:54:58 +01:00
Sam Gross
0d7d29fa57
Enable caching allocator for CUDA pinned memory ( #275 )
...
Also add binding for CUDA "sleep" kernel
2016-12-02 01:33:56 -05:00
Adam Paszke
ebc70f7919
Look for libcudart in default CUDA installation paths ( #195 )
2016-11-02 19:36:10 -04:00
Sam Gross
ad5fdef6ac
Make every user-visible Tensor have a Storage ( #179 )
2016-10-31 12:12:22 -04:00
Sam Gross
79ead42ade
Add CUDA Stream and Event API ( #133 )
2016-10-18 12:15:57 -04:00
Sam Gross
8d39fb4094
Use new THC API for device allocator
2016-10-17 09:35:41 -07:00
Sam Gross
ee14cf9438
Add support for pinned memory: ( #127 )
...
torch.Storage/Tensor.pin_memory()
torch.Storage/Tensor.is_pinned()
2016-10-15 18:38:26 -04:00
Sam Gross
c20828478e
Update Module.cpp for THC changes
2016-09-30 11:13:14 -07:00
Adam Paszke
3f7ab95890
Finish implementation of prng related functions
2016-09-29 11:33:25 -07:00
Sam Gross
4e9f0a8255
Use CUDA caching allocator
2016-09-26 13:12:39 -07:00
Adam Paszke
06ab3f962f
Refactor _C extension to export some utilities
2016-09-21 08:36:54 -07:00
Adam Paszke
3ea1da3b2c
Minor fix in CUDA module
2016-09-14 11:09:03 -04:00
soumith
1f2695e875
adding cuda driver check functions for runtime checking
2016-09-13 10:34:13 -07:00
Adam Paszke
8d933cbfc4
Fixes for OS X
2016-08-22 22:45:35 -04:00
Adam Paszke
12bed8dc0d
Add CUDA device selection
2016-08-12 07:46:46 -07:00
Adam Paszke
92e983a489
Fixes for Linux and new cutorch
2016-08-02 09:20:18 -07:00
Adam Paszke
c574295012
Various fixes
2016-07-19 10:45:59 -04:00
Adam Paszke
3a44259b32
Add support for CUDA
2016-07-19 10:45:59 -04:00
