-
Hi there,
I'm new to quantization. From my understanding, "8da4w" means that the weights are pre-quantized to 4 bits, and the activations are quantized to 8 bits at runtime. Following this, the GEM…
-
I am currently learning about the example llama2 by following the instructions provided in the [README](https://github.com/pytorch/executorch/blob/main/examples/models/llama2/README.md). In this examp…
-
### Summary
This project is undertaken as part of my summer internship at Quansight Labs, mentored by @izaid and @rgommers.
I would like to propose extending the current COO sparse array support in …
-
-
So recently we had a user who wanted to model multiple assessments per visit per subject. We currently cannot model this adequately with the existing covariance structure framework, because:
1) we …
-
We should use the `COO` sparse format when we construct a sparse matrix in several steps (add each transition at a time). In the `COO` format (see `SparseMatrixCOO` from [LuxurySparse.jl](https://gith…
-
As AA has been refactored in a class, we should make sure to catch a sneaky bug that can make AA fail silently.
I've already stumbled upon cases where the residual matrix is ill-conditioned, which ma…
-
To integrate various specialized processors such as Tensor Processing Units (TPUs), Language Processing Units (LPUs), Graphics Processing Units (GPUs), and others into the Cyclops-64 architecture, we …
-
I am running into an issue with JAX where after successive vmaps one gigantic matrix is created (or at least pre-allocated). In my case I am computing kernel matrices:
```python
import jax
import…
-
this should sped up computations, especially matrix inversions?
dswah updated
7 years ago