-
The data download scripts are segfaulting for me on an Ubuntu p3.8x machine on aws
## Description
```
python3 datasets/dataset_setup.py \
--data_dir=~/data \
--librispeech
```
Both libris…
-
### Bug description
I am using the new `configure_model` method to initialize my models. However when I try to resume training from the saved checkpoints. I am getting an `Unexpected key(s) in stat…
-
## Motivation
This is somewhat related to https://github.com/pytorch/rl/issues/849. I realize TorchRL is quite new and I expect the documentation and examples will improve over time, but I believe th…
-
Flax, optax, jaxopt to my understanding are only required for neural network based OT.
There may be situations whereby these are not needed. For example if some users prefer to use haiku instead o…
-
Thanks for your great job,
Could you please upload a docker image if you have one? I'm struggling with the environment....
Thanks for your future reply!
-
I apologise in advance as I may have misunderstood something obvious, as I haven't used probabilistic ODE solvers before and am coming from using Diffrax.
If one wants to use simulate_terminal_val…
-
Hi,
I've been using a Sinkhorn loss in my code, and it was quite fast on my laptop's CPU. But when moving to a GPU cluster, my code suddenly ran more than 10 times slower. This is not the case when…
-
### Describe the bug
I tried simple code example in Google's Colab and there is no response or error. It just hangs.
```
import openllm
client = openllm.client.HTTPClient("http://localhost:3000…
-
Hi all,
Thought I would create a separate issue for this. Essentially, we want to use the `_mft` in the `GaussianPropagator` in the long term to avoid sampling difficulties and increase the speed. As…
-
### What happened + What you expected to happen
1. I use `add_column` to add a empty dict `meta` to dataset and then use `map` to update the `meta` dict, but get a wrong dataset.
```python
d…