-
This could also be an issue for other operators.
It should be possible to create them even after moving data to cuda.
Currently, you would have to move to CPU, create, move to GPU again..
-
After training the model on a small portion of the data (a couple hundred thousand frames as suggested), and then applied the model to the rest of the dataset. After doing this, `generate_trajectory_p…
-
There are two key components, `GFlowNet` (in `src/gfn/gflownet`) and `Sampler` (in `src/gfn/samplers.py`).
I think `Sampler` is a redundant class, and we could get rid of it by incorporating the sa…
-
Here I note several issues in the current `ReplayBuffer`, including the points raised by @younik in #193.
1. **Types of objects**, (#193): Currently, `ReplayBuffer` can take `Trajectories`, `Transi…
-
I want to use 3d slam in my ardupilot_ros drone, and I add the depth camera in the SDF file. When I use 3D slam, the viewing angle in the z-axis direction is too large so that it seems to make the gro…
-
The current index files available on GDAC https://data-argo.ifremer.fr (*index*.txt) are not homogeneous.
A richer and homogeneous series of index files should be managed.
- Dedicated directory fo…
-
I trained a 128x64 model (with sparse=False); if I record the ~39 *denoiser outputs* during the course of a single sampling loop and compute their mse loss (*not* weighted by lambda_sigma because this…
-
Thanks for the great work! Are there any tips for training with the improved contrastive divergence objective? I'm trying to build a multi-modal robotic manipulation model that takes in videos and tex…
-
First of all: Thank you for your work.
The issue is in "Studying trajectories: Ephem objects -->"
🐞 **TypeError: time_range() takes 1 positional argument but 2 were given**
The following cod…
-
Hi ! Thank you for your great work. I notice that in the trajectory generation part. you set the initial position, velocity and accelerate to 0;
```
int GetNextCandidateTrajectory(
Ra…