brrrrrm

19 ms·

1. 241.▲
   by brrrrrm 4y ago

   > huge for the progress of WASM HUGE! 4x speedup for anything parallelizable. This should bring around ~45GFlops out of the box in the iPhone browser, which isn't much but easily enough for basic neural networks (think object track
2. 242.▲
   United States National Radio Quiet Zone(en.wikipedia.org)

   8 points by brrrrrm 4y ago|4 comments
3. 243.▲
   by brrrrrm 4y ago

   I believe they mean exploit the arbitrage opportunity to the point that the opportunity ceases to exist
4. 244.▲
   GitHub Repo Named “Eeeeeeeeeeeeeeeeeeeeeeee ”(github.com)

   2 points by brrrrrm 4y ago|0 comments
5. 245.▲
   Better Bindings for Python(github.com)

   1 points by brrrrrm 4y ago|0 comments
6. 246.▲
   by brrrrrm 4y ago

   My claim is that it’s not easy, not impossible. There’s little incentive to hack in JavaScript or maintain a Pyodide compatible build. The 4gb limit isn’t a technical limitation, just a standards thing (it could change easily).
7. 247.▲
   by brrrrrm 4y ago

   the issue right now with Python support in WASM (at least for machine learning, the main driver of the language) is that Python is largely a wrapper language and none the utilities that make it so powerful (numpy, PyTorch, JAX) work particu
8. 248.▲
   by brrrrrm 4y ago

   This is a really cool effort! I’m curious, where do the bottlenecks live right now?
9. 249.▲
   by brrrrrm 4y ago

   There’s a C embed example in the examples folder
10. 250.▲
    WebAssembly SIMD Now Available in Safari Technology Preview(developer.apple.com)

    2 points by brrrrrm 4y ago|0 comments
11. 251.▲
    by brrrrrm 4y ago

    > If in the near future JS also reaches good support for scientific/ml libraries it seems like entering that ecosystem is an obvious advantage Fully agree with this sentiment. To that end, I’ve been contributing to Shumai (linked b
12. 252.▲
    by brrrrrm 4y ago

    Storage isn’t much of an issue these days. The bottleneck is computing values from RAM, fp8 can improve this drastically
13. 253.▲
    Sizing Up Neural Networks(jott.live)

    1 points by brrrrrm 4y ago|0 comments
14. 254.▲
    by brrrrrm 4y ago

    Mentioned in the post, Cornell provides some really cool interactive data: - fine grained map of the wages/% making livable wage: https://blogs.cornell.edu/livingwage/ - likelihood of making livable wage based on
15. 255.▲
    by brrrrrm 4y ago

    without using intrinsics? `-O3 -march=skylake-avx512 -mprefer-vector-width=512`
16. 256.▲
    by brrrrrm 4y ago

    well, AMX does asynchronous outer products from byte-addressable buffers with accumulation. it's not like tensor cores or intel's AMX where the silicon literally only does a matrix multiplication. you can easily do efficient slid
17. 257.▲
    by brrrrrm 4y ago

    typically if it's available, compilers will use the avx512 register file. This means you'll see things like xmm25 and ymm25 (128 and 256 bit registers) and those are avx512 only. However, compilers using 512-wide instructions is
18. 258.▲
    by brrrrrm 4y ago

    I don't see throughput information on Intel's AMX, but their VNNI has information: https://www.intel.com/content/www/us/en/docs/intrinsics-guid... 0.5 cycles per instruction max, 3.7GHz cl
19. 259.▲
    by brrrrrm 4y ago

    I believe the vim.wasm is pre-compiled and checked in as a "binary." The javascript written is a runtime wrapper
20. 260.▲
    by brrrrrm 4y ago

    author here! I grabbed them from here IIRC https://emojipedia.org/apple/ (there are a bunch of sites that have these) edit: yep - predicting the original based on the augmented version (as a hack to generalize for unse
21. 261.▲
    by brrrrrm 4y ago

    > It's really a shame that data science, ML, and notebooks are so wrapped up in it. Otherwise we could jettison the whole thing into space Although I personally feel Python has its place, I contribute to a project that hopes to dive
22. 262.▲
    by brrrrrm 4y ago

    I feel similarly. However, Python has - reference counted objects with predictable scoped destruction. this keeps the maintenance of huge resources quite simple and memory usage steady - operator overloading (infix, like +/*-) I real
23. 263.▲
    Rendering Images with Emoji in Real Time(jott.live)

    4 points by brrrrrm 4y ago|0 comments
24. 264.▲
    by brrrrrm 4y ago

    I’ve found javascript is much better than wasm for lots of this stuff https://bwasti.github.io/mebm/
25. 265.▲
    by brrrrrm 4y ago

    if everyone on earth shared a memory space and every person had 4 gigabytes of byte addressable storage, we'd be at 65 bits.
26. 266.▲
    by brrrrrm 4y ago

    does this support any form of FFI? It'd be nice if users could shim in lightweight APIs that clone libraries like numpy/pytorch -- it'd immediately make some machine learning super portable!
27. 267.▲
    by brrrrrm 4y ago

    ChatGPT uses a modified version of GPT3 (details are sparse, but here are some: https://beta.openai.com/docs/model-index-for-researchers ). There’s a logbook associated with the training of OPT (similar size to GPT-3):
28. 268.▲
    by brrrrrm 4y ago

    edit: TorchInductor calls into GCC/Triton depending. I mistook it for the various backends TorchDynamo supports (including TVM)
29. 269.▲
    Node-calls-Python: call Python from Node.js(github.com)

    6 points by brrrrrm 4y ago|0 comments
30. 270.▲
    by brrrrrm 4y ago

    Some context/history: For compiler people reading this, a lot of common compiler terms have been entirely reinvented in the context of machine learning frameworks. An ML "graph" refers almost exactly to the dataflow graph (D