Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Frozen parameters in GaussianFourierProjection

See original GitHub issue

Hi, just a beginner with diffusion models and have been using your implementations as reference. I have a question about this class

Why is requires_grad set to false in the weight parameter? Won’t this mean, during training, the noise level embeddings won’t be updated?

Thanks!

Issue Analytics

State:
Created a year ago
Comments:5 (5 by maintainers)

Top GitHub Comments

1reaction

vvvm23commented, Aug 31, 2022

Okay thank you @patrickvonplaten ! That explanation makes a lot of sense~

0reactions

patrickvonplatencommented, Aug 30, 2022

Hey @vvvm23,

sinusoidal position features like GaussianFourierProjection don’t need training because every embedding already has a distinctly different vector that the model can use a “cue” to know what time position has been passed to it.

If one wants to train position embedding vectors (or time embedding vectors here), one can just randomly initialize such a vector and let the model learn it. If however we use sinusoidal embeddings, there is no need to learn it

Top Results From Across the Web

Frozen Gaussian approximation for high frequency wave ...

In this paper, based on Bloch decomposition and asymptotic analysis in the phase space, we derive the frozen Gaussian approximation for ...

Nonadiabatic Quantum Dynamics with Frozen-Width Gaussians

Abstract We review techniques for simulating fully quantum nonadiabatic dynamics using the frozen-width moving Gaussian basis functions to ...

Density functional theory of freezing - KU ScholarWorks

more convienently evaluated by decomposing the Gaussian density in a Fourier series since the integral becomes a sum over reciprocal lattice vectors and ......

frozen Gaussian approximation for high frequency wave ...

Abstract. We propose the frozen Gaussian approximation for computation of high frequency wave propagation. This method approximates the ...

An error bound for the time-sliced thawed Gaussian ...

We study the time-sliced thawed Gaussian propagation method, which was recently proposed for solving the time-dependent Schrödinger equation ...