-
Hello,
I really appreciate what you guys have been doing, for my first time with these splats / nerfs, this is the first time you get really good instructions and everything works out of box, allmo…
-
Hi -
I was looking to retrieve forces from Orca logfiles and noticed https://cclib.github.io/data_notes.html#grads describes `grads` as corresponding to forces. However, looking at the parser - ht…
-
### Background
We need to support grads in making informed decisions about their futures. Group messaging and discussions are limited in supporting this, as we can't write specific messages that ta…
-
It appears that grads and grads_abs may not be getting cleared correctly between calls of densify_and_prune(). They seem to double in size each time, eventually leading to out of memory errors. As suc…
-
Related to #206 and partly to #66, a feature request: allow sparse gradients for getindex, as [pytorch does](https://pytorch.org/docs/stable/nn.html?highlight=embedding#torch.nn.functional.embedding) …
-
Dear torchtitan team, I have a question regarding gradient norm clipping when using pipeline parallelism (PP) potentially combined with `FSDP/DP/TP`.
For simplicity, let's assume each process/GPU h…
-
### 🚀 The feature, motivation and pitch
I have a python code segment related to a deep RL algorithm where it calculates the second order optimization and second derivative with Hessian matrix and f…
-
I met the following error when testing sgd
```
@data.setter
def data(self, value):
assert isinstance(value, Tensor)
> assert value.dtype == self.dtype, "%s %s" % (
…
-
### Background
Classes run best when Grads are involved. How do we engender a community where grads come back?
- Get list of engaged grads in each region - host collective discussion
- Share goals…
-
Ideas from @blakehechtman:
1. make [the default value](https://github.com/google/jax/blob/6cb8737c1a1fb3f4d647f5354f7c31d7971a440d/jax/test_util.py#L82) of EPS in `check_grads` depend on dtype, rathe…