-
* The `collect` function recursively `concat`s on a tree, making the time complexity quadratic. This can be avoided using difference lists.
* The `do_label` step builds some unnecessary intermediate …
-
The current naive implementation is $O(n^2)$, whereas [Bentley–Ottmann](https://en.wikipedia.org/wiki/Bentley%E2%80%93Ottmann_algorithm) is $O((n+k)\log n)$. Plans to implement?
-
As best as I can tell, the most efficient way to get diag(JtJ) is to do jnp.square(primals).T @ jnp.square(cotangents), but this seems to be difficult to cleanly implement in JAX.
I have an utterly…
ywrt updated
3 months ago
-
I have installed [obsidian-docker](https://docs.linuxserver.io/images/docker-obsidian) and I am using livesync plugin on it. Whenever the sync starts on it, it takes too much RAM and crashes the raspb…
-
First of all, great project and very useful. It has a lot of promise and probably should be integrated with mistral-rs and other rust llm projects.
I noticed that it doesn't follow llama.cpp's meth…
-
Current implementation is not friendly for the development of line search and will need some refactoring.
-
I am having trouble understanding and finding the correct way to use sumMap efficiently both
ön materialized input and ordered (limited output). Given the following;
```sql
create table log (
…
bjne updated
3 months ago
-
Hello, I have recently implemented a cross attention application with multi-modal fusion, but because the image resolution is too large, cuda OOM occurs when calculating q and k, so I found your paper…
-
The leading quantum computing SDK/library is Qiskit by IBM. However it is a Python library. There are some undocumented ways to use C++ with Qiskit, but generally it is for Python. Preferably the back…
-
I occasionally run into issues when manipulating AxisArray objects when I fail to properly break the link between the incoming message and the newly generated output. A message will get manipulated in…