cloudhan

7 ms·

1.
by cloudhan 1y ago
> The official project language is American English with ISO 8601 dates and metric units. Very good taste for an American.
2.
by cloudhan 2y ago
NPU folks offen time say > it's not enough time to get new silicon designs specifically for <blahblah> Where blahblah stands for a model architecture that has caused a paradigm shift. When you need a new silicon for a new mode
3.
by cloudhan 2y ago
OK, I am one of the developers in onnxruntime team. Perviously working on ROCm EP now has been transfered to QNN EP. The following is purely devrant and the opinions are mine. So ROCm already sucks whereas QNN sucks even harder! The conclus
4.
by cloudhan 2y ago
Yes.
5.
by cloudhan 2y ago
This reminds me of Weight Agnostic Neural Networks https://weightagnostic.github.io/
6.
by cloudhan 2y ago
Very unfriendly. The symbolic library (type of activations) requires a branching at the very core of the kernel. GPU will need to serialized on these operations warp-wise. To optimize, you might want to do a scan operation beforehand and di
7.
by cloudhan 3y ago
You run or your code runs, choose one and choose it wisely ;)
8.
by cloudhan 3y ago
The memory is too small to be useful nowadays.
9.
by cloudhan 3y ago
At the time of the dev of the EP, the tooling is not that good as current. I have imagined a pipeline that HLSL compiles down to DXIL then go through spirv cross and then target wide variety of mobile devices with opencl runtime. But they a
10.
by cloudhan 3y ago
Not exactly, both cl and glsl can be aot, but the runtime will be limited to some newer version and then the market coverage will be niche, those vendor are so lazy on updating the driver and fixing the compiler bugs...
11.
by cloudhan 3y ago
circle lang is also very worth to check out.
12.
by cloudhan 3y ago
Thoes "existing libraries" does not necessary mean stdc++, but some parallel primitive, and are essential to performance portability. For example, cub for scan and reduction, cutlass for dense linear algebra[1]. > I think the d
13.
by cloudhan 3y ago
I am long sought after a CUDA or HIP compiler that target SPIR-V or DXIL. So that we can compile all thoes neural network kernels to almost all compute devices. The requirements are, 1. extend on C++17 this means template meta programming
14.
by cloudhan 3y ago
Mixtral-8x7B support --> Support new model https://github.com/stanford-futuredata/megablocks/pull/45
15.
by cloudhan 3y ago
Might be the training code related with the model https://github.com/mistralai/megablocks-public/tree/pstock/m...
16.
by cloudhan 3y ago
You might also want to watch on helix, if PR 8675 is merged, then scheme will be the extension langauge, then we get the best of both world. Modal editing and scheme =) PR 8675: https://github.com/helix-editor/helix&#x2
17.
by cloudhan 3y ago
onnxruntime is written in C++. So it is not misleading =)
18.
by cloudhan 3y ago
This papar is really interesting, especially so when you scroll to the Reference part ;)
19.
by cloudhan 5y ago
I think you might be interesting in this interactive papar: https://willcrichton.net/nota/
20.
by cloudhan 5y ago
Then we need a powerpoint virtual machine[1]. [1] https://www.andrew.cmu.edu/user/twildenh/PowerPointTM/Paper....