Building on HF

PZ

philipp-zettl

https://blog.godesteem.de

philsupertramp

AI & ML interests

NLP/CV/Multimodal learning

Recent Activity

upvoted a paper about 5 hours ago

ZAYA1-8B Technical Report

liked a model about 5 hours ago

Zyphra/ZAYA1-8B

updated a Space 3 days ago

philipp-zettl/vecdb-wasm

View all activity

Organizations

My [ml-intern]( smolagents/ml-intern) assisted me yesterday to build a vector DB that runs in your browser
philipp-zettl/vecdb-wasm

To demonstrate the capabilities of this engine, the intern then went ahead and implemented a fully offline document RAG application
philipp-zettl/academic-copilot

And that was just one of my interns working yesterday 😄

Post

2616

I've been cooking something neat over the past weeks 👨‍🍳

We all know that training LLMs requires a lot of resources and especially a lot of compute in form of GPUs, or is super slow and inefficient when done on CPUs.

The big players use giant clusters of Nvidia H100s.
But if I look at the profiles of my fellow home brewers, all we can get our hands on are those pesky consumer RTX's. If you're lucky you got yourself a 5080 with 16GB VRAM or something.

To be frank, I don't have that 1.3k disposable cash laying around ¯\_(ツ)_/¯
But I can write rust and like building ML libraries.

So I asked myself the question(s):
- can I train SMLs at home on my hardware?
- How hard can it be to build a ML library that can stream data between RAM and VRAM on demand, like llama.cpp's unified memory feature [^1]?
- how hard can it be to implement bf16 support?

The answers are wild, trust me!

Image 1: Metrics form last nights build on my "tiny" RTX 2060 (6 GB VRAM)
Image 2: Metrics from my most recent build on my RTX 4070 Laptop (8GB VRAM)

The majority of my time went into the shared memory, but it's stable and I'm very excited!
Here some debug logs, a la "trust me bro"

----
Currently available: 1112735744, attempting to reclaim: 1073741824
--- VRAM STATE [backward pass] ---
Driver Used:    6744 MB / 7805 MB
Data on GPU:    1641 MB
Grads on GPU:   3459 MB
CPU Offloaded: 18230 MB
---------------------------------
Currently available: 1079181312, attempting to reclaim: 1073741824
--- VRAM STATE [backward pass] ---
Driver Used:    6776 MB / 7805 MB
Data on GPU:    1561 MB
Grads on GPU:   3279 MB
CPU Offloaded: 18590 MB
-----------------------------

Final models get exported in safetensors format and are compatible with PyTorch and transformers, for accessibility.

- [^1]: https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md#unified-memory