-
## 🚀 Feature
Implement GPU INT8 matrix multiplication in PyTorch.
## Motivation
In order to save time by using less GPU memory per data (hence, being able to use bigger batch sizes), I …
-
```
dev` profile [unoptimized + debuginfo] target(s) in 3m 33s
Running `D:/rust-target\debug\gpt-burn.exe train --context-length 128 --n-layers 12 --n-heads 12 --d-model 768 --batch-size 128 --…
-
Would it be feasible at all to add PyTorch support as a substrate?
Besides the obvious use case, this would enable researchers to implement framework-agnostic probabilistic algorithms that use a Nu…
-
Leaf, and now Spearow, always promised OpenCL but has yet to deliver. I think that's a shame!
There are plenty of great folks working on ML experiments for Rust which feature OpenCL or frameworks t…
ghost updated
6 years ago
-
Goal:
- Create a CharacterCard class that extends HorseVariable. Implement the `apply_gradients` and `summary` functions. SillyTavern, Risu, Agnai, and CharacterAI cards are preferred.
- Add a test …
-
```julia
using OrdinaryDiffEq, RecursiveArrayTools, LinearAlgebra, Test, SparseArrays, SparseDiffTools, Sundials
# Define the constants for the PDE
const α₂ = 1.0
const α₃ = 1.0
const β₁ = 1.0
…
-
## Bug prerequisites
1. When ACTS v21.1.1 is built with flags:
```bash
cmake -Wno-dev -DCMAKE_INSTALL_PREFIX=/home/user/acts/acts-v21.1.1
-DCMAKE_CXX_STANDARD=17
-DACTS_B…
-
Hi,
I was considering some ideas with EFP in Pytorch but I get from EFPSet.batch_compute
`TypeError: arg is not one of numpy.ndarray, list, or fastjet.PseudoJet`
Any chance to support also torch.…
-
High-level API for model training similar to PyTorch Lightning.
One advantage over Python is that we don't need a separate module type to add methods to a model class (e.g. train_step() and similar…
-
Comparison with eclipse 100 seems to indicate that injectors should be treated similar to producers, thus simply be assigned the properties of the perforation grid cells ...