-
Hi Tim!
Many thanks for this awesome repo and your [paper](http://arxiv.org/abs/1907.04840).
It is always cool when someone tries DL actually useful, accessible and more efficient!
We are build…
-
## 🚀 Feature
I think because of static linking, we can use PyTorch with CUDA without installing CUDA to the host system, or CUDA version on the system and PyTorch can differ.
However, when i…
ozen updated
11 months ago
-
-
### Description of bug
I am running the following command "spades.py --rna -s mouse1_mRNA.fastq -o mouse1_spades -m 50 --phred-offset 33" getting error like this == Error == system call for: "['/hom…
-
Please publish a rockspec so that the library can be installed via luarocks (and used as a dependancy!)
-
Reproducer on branch `105-compilation-time-with-vmap`
```
git checkout 105-compilation-time-with-vmap
export JAX_PLATFORMS=cpu,ipu
export TF_POPLAR_FLAGS="--show_progress_bar=true"
pytest -k te…
-
I was trying to run the default colab notebook but I keep getting the following error:
from trl import SFTTrainer
from transformers import TrainingArguments, DataCollatorForSeq2Seq
from unsloth i…
-
I just noticed that `Conv1d` isn't made visible in https://github.com/graphcore-research/unit-scaling/pull/67
So for now one needs to e.g.
```python
from unit_scaling._modules import Conv1d
```
i…
-
Continuing from https://github.com/franko/graph-toolkit/issues/3#issuecomment-68569447
> the current version of the library won't work in Lua > 5.1; for one, it uses C API functions that were removed…
-
I am trying to reimplement u-MuP in JAX and using this repo as a reference. I cannot figure out why Adam and SGD use different lr scaling factors: $\frac{1}{\sqrt{\text{fan-in}}}$ for [Adam](https://g…