uwsampl / pytorch Goto Github PK
View Code? Open in Web Editor NEWDynamic tensor rematerialization, implemented as a pytorch fork
License: Other
Dynamic tensor rematerialization, implemented as a pytorch fork
License: Other
Steps to reproduce the behavior:
Please copy and paste the output from our
environment collection script
(or fill out the checklist below manually).
You can get the script and run it with:
wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py
conda
, pip
, source):import torch
from torch import tensor
a = torch.randn(4, 4).checkpoint()
a = torch.tensor([-1, -2, 3], dtype=torch.int8).checkpoint()
torch.bitwise_and(a, a)
ultimately this is due to bitwise_and initiate a tensor and call mutate on it internally.
two ways to fix this
0: I can allow checkpoint_function to take in raw tensor and mutate them. this I can do quite easily but is unsafe - while I can reflect change on cptensor back to raw tensor, I cant do the reverse without modifying the old tensors.
1: I can change bitwise_and. the obvious problem is that this is non-extendable. But why can bitwise_and initialize tensor? it must know whether the input is sparse/dense and initalize likewise. we can find this hopefully generic code and add to it.
I will go with route 1 for now.
Steps to reproduce the behavior:
Please copy and paste the output from our
environment collection script
(or fill out the checklist below manually).
You can get the script and run it with:
wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py
conda
, pip
, source):get a vector of cptc for clear_checkpointpool, instead of reusing pool.
tensors in pool maybe incomplete and out of order due to eviction, so getting a seprate pool will fix this.
my system friend tell me to use https://github.com/oneapi-src/oneTBB
we can also roll our own producer-consumer.
Steps to reproduce the behavior:
Please copy and paste the output from our
environment collection script
(or fill out the checklist below manually).
You can get the script and run it with:
wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py
conda
, pip
, source):Reformer
NAS inference
http://pyro.ai/examples/dmm.html
https://github.com/facebookresearch/clevr-iep
this will bring our total model count into 12, which we can layout 3*4 manner in the paper.
However, I am also thinking of Meta Learning with:
CNN
LSTM.
maybe we should replace a normal CNN/LSTM example with those.
MAML itself use CNN.
https://github.com/twitter/meta-learning-lstm
https://github.com/cflamant/neural-stack
Adaptive Computation Time
https://github.com/kentsommer/pytorch-value-iteration-networks/blob/master/model.py
https://github.com/ikostrikov/pytorch-meta-optimizer/blob/master/meta_optimizer.py
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.