JayMan91

repos

gists

followers

following

JayMan91/aaai_predit_then_optimize 5

Code release for AAAI 2020 paper "Smart Predict-and-Optimize for Hard Combinatorial Optimization Problems"

JayMan91/NeurIPSIntopt 5

JayMan91/aaai_melding_code 0

Code the AAAI 2019 paper "Melding the Data-Decisions Pipeline: Decision-Focused Learning for Combinatorial Optimization"

JayMan91/academicpages.github.io 0

Github Pages template for academic personal websites, forked from mmistakes/minimal-mistakes

JayMan91/attention-learn-to-route 0

Attention based model for learning to solve different routing problems

JayMan91/blackbox-backprop 0

Torch modules that wrap blackbox combinatorial solvers according to the method presented in "Differentiating Blackbox Combinatorial Solvers"

JayMan91/Combinatorial-Opt 0

JayMan91/dask-tutorial 0

Dask tutorial

JayMan91/deep-limits 0

Repo for a paper about constructing priors on very deep models.

JayMan91/e2e-model-learning 0

Task-based end-to-end model learning in stochastic optimization

startedeambutu/snail-pytorch

started time in an hour

startedsamcohen16/Aligning-Time-Series

started time in 6 hours

startedgeomstats/applications

started time in a day

issue openedJayMan91/NeurIPSIntopt

EnergyScheduling experiments not working

Hello, running PYTHONPATH=. python3 ./exp_run.py in the EnergyScheduling folder leads to this:

[..]
Epoch[4/11], loss(train):384595.25 @ 2021-01-19 19:36:28 
Epoch[4/12], loss(train):349502.72 @ 2021-01-19 19:37:53 
Epoch[4/13], loss(train):309425.31 @ 2021-01-19 19:39:19 
Epoch[4/14], loss(train):358522.38 @ 2021-01-19 19:40:40 
Epoch[4/15], loss(train):320667.47 @ 2021-01-19 19:41:59 
Epoch[4/16], loss(train):261996.72 @ 2021-01-19 19:43:22 
Epoch[4/17], loss(train):303735.38 @ 2021-01-19 19:44:43 
Epoch[4/18], loss(train):358828.66 @ 2021-01-19 19:46:02 
Epoch[4/19], loss(train):361545.69 @ 2021-01-19 19:47:22 
Epoch[4/20], loss(train):262988.81 @ 2021-01-19 19:48:42 
Epoch[4/21], loss(train):262221.91 @ 2021-01-19 19:50:02 
Epoch[4/22], loss(train):478453.25 @ 2021-01-19 19:51:23 
Epoch[4/23], loss(train):354604.09 @ 2021-01-19 19:52:45 
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Epoch[1/1] fwd pass not solved
Traceback (most recent call last):
  File "./exp_run.py", line 71, in <module>
    clf.fit(X_1gtrain,y_train,X_test= X_1gtest,y_test= y_test)
  File "../../Interior/intopt_energy_mlp.py", line 627, in fit
    i+1, loss.item(),datetime.datetime.now() ))
UnboundLocalError: local variable 'loss' referenced before assignment

It's pretty easy to fix, I can do a pull request later today.

created time in 4 days

issue openedJayMan91/NeurIPSIntopt

MultilayerRegression model

Hello, great work!

One question - the MultilayerRegression model (defined in https://github.com/JayMan91/NeurIPSIntopt/blob/c963927a3e40934312ddfdf400f10eb506acc7dc/Interior/intopt_energy_mlp.py#L34) seems to be defined as a sequence of linear layers (and thus still a linear model) without any non-linearity.

Can you confirm that's correct, and that is the model you used for producing the results in your paper?

created time in 4 days

startedJayMan91/NeurIPSIntopt

started time in 5 days

startedJayMan91/NeurIPSIntopt

started time in 5 days

startedgoogle/flax

started time in 7 days

starteddeepmind/dm-haiku

started time in 7 days

starteddfm/extending-jax

started time in 12 days