A tracing JIT for PyTorch
Перейти к файлу
Nuno Lopes 2c36cc5024 stride inference 2021-12-11 12:58:16 +00:00
autogen stride inference 2021-12-11 12:58:16 +00:00
backends updates for pytorch 1.10+ 2021-10-27 22:27:13 +01:00
benchmarks add microbenchmark with control-flow 2021-11-12 17:21:51 +00:00
scripts stride inference 2021-12-11 12:58:16 +00:00
tests fix bug with missing event of 'tensor.data = tensor2' 2021-10-19 15:14:38 +01:00
torchy stubs 2021-03-25 20:18:06 +00:00
.gitattributes run unit tests with both backends 2021-09-08 18:21:01 +01:00
.gitignore stride inference 2021-12-11 12:58:16 +00:00
LICENSE Initial commit 2021-03-25 20:15:04 +00:00
PYTORCH_COMMIT working on better type inference (WIP) 2021-11-28 23:16:54 +00:00
README.md make the example work (almost) :D 2021-04-16 18:14:12 +01:00
UPGRADE_PYTORCH.md dtype infer updates 2021-11-30 20:26:45 +00:00
autograd.cpp add basic strides inference support 2021-10-28 19:39:55 +01:00
collect-stats.sh stride inference 2021-12-07 15:41:28 +00:00
config.h stash stats 2021-05-28 14:30:09 +01:00
disable_shape_infer.patch working on better type inference (WIP) 2021-11-28 23:16:54 +00:00
dispatch.h add autograd fallback handler 2021-04-27 18:27:03 +01:00
dot.sh stats: add dot printer 2021-06-02 16:33:21 +01:00
gen.py stride inference 2021-12-11 12:58:16 +00:00
init.cpp add tape 2021-03-29 19:22:44 +01:00
ops.cpp stub a torchscript backend 2021-09-03 15:30:46 +01:00
ops.h stub a torchscript backend 2021-09-03 15:30:46 +01:00
pytorch.patch fix instanceof(foo, *Tensor) where we would always give false 2021-10-05 18:04:50 +01:00
run-benchmarks.php also benchmark microbench with torchscript 2021-11-05 18:18:00 +00:00
run-tests.sh add unit test framework 2021-04-29 17:20:42 +01:00
setup.py misc simplifications 2021-10-24 15:50:45 +01:00
shape_inference.h stride inference 2021-12-11 12:58:16 +00:00
stats.cpp add basic support for autograd 2021-10-28 13:52:56 +01:00
stats.h add basic support for autograd 2021-10-28 13:52:56 +01:00
stopwatch.cpp stats: collect run-time per trace 2021-06-01 12:20:04 +01:00
stopwatch.h stats: collect run-time per trace 2021-06-01 12:20:04 +01:00
strides_inference.h stride inference 2021-12-11 12:58:16 +00:00
tensor.cpp stride inference 2021-12-11 12:58:16 +00:00
tensor.h add basic strides inference support 2021-10-28 19:39:55 +01:00
trace.cpp stride inference 2021-12-11 12:58:16 +00:00
trace.h updates for pytorch 1.10+ 2021-10-27 22:27:13 +01:00
type_inference.h stride inference 2021-12-11 12:58:16 +00:00
typings_data.py stride inference 2021-12-11 12:58:16 +00:00
utils.h add caching mechanism for compiled traces 2021-09-24 18:02:01 +01:00

README.md

Torchy

A tracing JIT for PyTorch. WIP; don't use.

Install

$ python setup.py install

Run

Torchy shouldn't require any change beyond adding a call to torchy.enable(). Example:

import torch
import torchy

torchy.enable()

x = torch.tensor(((3.,2.), (4.,5.)))
y = torch.tensor(((5.,6.), (7.,1.)))

w = x.add(x)
z = x.add(y)
w = None  # w not computed
print(z)