Comments (11)
from pytorch-ts.
from pytorch-ts.
can you try with the 0.7.0 branch and updated gluonts?
from pytorch-ts.
from pytorch-ts.
pip install diffusers
perhaps i forgot to add it to the setup.py
from pytorch-ts.
from pytorch-ts.
Hi @kashif, I want to use DeepVAR model which is not supported in 0.7.0 branch. If I use the default branch, I get the error of "Reached maximum number of idle transformation calls". Do you have any suggestions? Thanks a lot!
from pytorch-ts.
Version 0.7.0 does not have the updated model creation for the TimeGrad example. You need to import a scheduler (the diffusion solver) from the diffusers library https://huggingface.co/docs/diffusers/api/schedulers/overview.
For the issue of "Reached maximum number of idle transformation calls." it's possible you do not have enough data, one solution is to just duplicate the data. Another issues is that if you're doing custom data you must ensure that the provided data is in the shape (input_size, timesteps) instead of (timesteps, input_size). It is possible that 0.7.0 has solved this issue as I have not encountered it since moving versions but I am using custom data.
Example on how to instantiate TimeGrad in 0.7.0
from diffusers import DEISMultistepScheduler
scheduler = DEISMultistepScheduler(
num_train_timesteps=150,
beta_end=0.1,
)
estimator = TimeGradEstimator(
input_size=int(dataset.metadata.feat_static_cat[0].cardinality),
hidden_size=64,
num_layers=2,
dropout_rate=0.1,
lags_seq=[1],
scheduler=scheduler,
num_inference_steps=150,
prediction_length=dataset.metadata.prediction_length,
context_length=dataset.metadata.prediction_length,
freq=dataset.metadata.freq,
scaling="mean",
trainer_kwargs=dict(max_epochs=200, accelerator="gpu", devices="1"),
)
from pytorch-ts.
@ProRedCat I get bad performance at version-0.7.0. Can you reproduce the similar performance recorded at timegrad-electricity notebook file?
from pytorch-ts.
Could be the solver you're using, DEISMultistepSchedular is a fast ODE solver but may perform worse than some other solvers.
I was able to get a lower score on Electricity of 0.018 with the DEISMultistepSchedular but the number of epochs was set to 200. What sort of performance numbers are you getting?
from pytorch-ts.
@ProRedCat I used the DDPMScheduler but I ran only 20 epochs. That might be the reason. I will try with DEISMultistepSchedular and large epochs. Thanks.
from pytorch-ts.
Related Issues (20)
- Branch: 0.7.0 - RuntimeError: Cannot serialize type diffusers.schedulers
- Run out of memory when I tried to run "Time-Grad-Electricity.ipynb" HOT 2
- Missing Trainer in version-0.7.0 HOT 1
- Enhancing Covariate Conditioning in TimeGrad HOT 1
- Multivariate-Flow-Solar:an error is reported when flow_type='MAF' HOT 1
- Reproducibility issue in TimeGrad with ver-0.7.0 HOT 8
- Inquiry about implementation of mean_wQuantileLoss and m_sum_mean_wQuantileLoss
- A question about the hyperparameter Settings of the model Time-Grad on both of Solar and Wikipedia datasets.
- Issue while runing the Readme
- can't generate dataset "pts_m5" HOT 5
- TypeError: `model` must be a `LightningModule` or `torch._dynamo.OptimizedModule`, got `TimeGradLightningModule`
- ValidationError: 1 validation error for PyTorchPredictorModel
- TypeError: PyTorchPredictor.__init__() got an unexpected keyword argument 'freq' HOT 14
- too many indices for array: array is 1-dimensional, but 2 were indexed
- Data imputation.
- TimeGrad Notebook version 0.7.0 -> predicts all nans HOT 5
- TimeGrad-electricity error
- Pytest pydantic throws an error
- Reproducing the results in "Multivariate Probabilistic Time Series Forecasting via Conditioned Normalizing Flows" in need of Parameters
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from pytorch-ts.