Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

github.com

Post reply on HN
Print view View original

21–30 of 33 posts

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#21

Post by AutoJanitor »

acmiyaguchi wrote: post #9

This feels like an AI agent doing it's own thing. The screenshot of this working is garble text ( https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b... ), and I'm skeptical of reasonable generation with a small hard-coded training corpus. And the linked devlog on youtube is quite bizzare too.

This is the text inference issues I was alluding to. We had several hurdles to overcome. 1 llms were trained on little endian. Mips for n64 is big endian. 2 we had python to c issues. 3 we had quantization issues. all being resolved. This is a tech demo to honor LOZ and also the code can be used for n64 devs to add ai style npcs in the future. So did we achieve it yes we are the first to do llm inference on n64. I am just trying to give you guys the proper video.

Scott

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#22

Post by AutoJanitor »

andrekandre wrote: post #20
AutoJanitor wrote: post #19

Earlier quoted context omitted.

We are going to use the gpu 128simd soon but it only has 4kb ram addressable so matmul offload in small chunks!

thats such really cool work; i wish i could get payed to do stuff like this, more power to you all ^^

I am doing this for 0 dollars. I am a self funded ai research lab. So when people diss me i get a little jaded, but then I remember I am doing cool stuff. Even if others don't see it. Thats enough for me!

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#23

Post by AutoJanitor »

mlaux wrote: post #5

I tried to build this but it's missing the weights.bin file and my computer is too weak to generate it. Can you add it to the repo?

Uploading weights.bin its really meant for you to generate your own llm but we are uplaoding it. They are ripping on it but they didnt check the code themselves. THis is a tech demo. its not about graphics its about the llm is inferring on the hardware lol.

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#24

Post by AutoJanitor »

danbolt wrote: post #15
gbnwl wrote: post #10

Earlier quoted context omitted.

You mean to tell me the included screenshot hasn't convinced you? https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...

I think the source code in the GitHub repo generates the ROM in the corresponding screenshots, but it seems quite barebones. It feels very much like it’s cobbled together from the libdragon examples directory. Or, they use hardware acceleration for the 2D sprites, but then write fixed-width text to the frambuffer with software rendering.

Partially correct. The value is not the game interface right now. Its proof you can do actual inference on an LLM the surprise I am developing is a bit bigger than this, just have to get the llm outputs right first!

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#26

Post by AutoJanitor »

Honest Limitations

    819K parameters. Responses are short and sometimes odd. That's expected at this scale with a small training corpus. The achievement is that it runs at all on this hardware.
    Context window is 64 tokens. Prompt + response must fit in 64 bytes.
    No memory between dialogs. The KV cache resets each conversation.
    Byte-level vocabulary. The model generates one ASCII character at a time.
Future Directions

These are things we're working toward — not current functionality:

    RSP microcode acceleration — the N64's RSP has 8-lane SIMD (VMULF/VMADH); offloading matmul would give an estimated 4–8× speedup over scalar VR4300
    Larger model — with the Expansion Pak (8MB total), a 6-layer model fits in RAM
    Richer training data — more diverse corpus = more coherent responses
    Real cartridge deployment — EverDrive compatibility, real hardware video coming
Why This Is Real

The VR4300 was designed for game physics, not transformer inference. Getting Q8.7 fixed-point attention, FFN, and softmax running stably at 93MHz required:

    Custom fixed-point softmax (bit-shift exponential to avoid overflow)
    Q8.7 accumulator arithmetic with saturation guards
    Soft-float compilation flag for float16 block scale decode
    Alignment-safe weight pointer arithmetic for the ROM DFS filesystem
The inference code is in nano_gpt.c. The training script is train_sophia_v5.py. Build it yourself and verify.

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#28

Post by AutoJanitor »

Ok I promised videos here is two. LLM had serious head issues with C and python x86 versus mips c. now coherent english. Phase two is chat interface so we can prompt without seeded prompts, check the code its real inference though!

The Emulator ---------------------------------------------- https://bottube.ai/watch/shFVLBT0kHY

The real iron! it runs faster on real iron! ---------------------------------------------- https://bottube.ai/watch/7GL90ftLqvh

The Rom image ---------------------------------------------- https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#29

Post by AutoJanitor »

great_psy wrote: post #2

Any place where we can test the llm output without loading it on n64? Curious of what we can get out of those constraints.

Ok I promised videos here is two. LLM had serious head issues with C and python x86 versus mips c. now coherent english. Phase two is chat interface so we can prompt without seeded prompts, check the code its real inference though! The Emulator ---------------------------------------------- https://bottube.ai/watch/shFVLBT0kHY

The real iron! it runs faster on real iron! ---------------------------------------------- https://bottube.ai/watch/7GL90ftLqvh

The Rom image ---------------------------------------------- https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...

reply

acmiyaguchi 19 hours ago | prev | next [–]

This feels like an AI agent doing it's own thing. The screenshot of this working is garble text (https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...), and I'm skeptical of reasonable generation with a small hard-coded training corpus. And the linked devlog on youtube is quite bizzare too.

Re: Happy Zelda's 40th first LLM running on N64 hardware (4MB RAM, 93MHz)

#30

Post by AutoJanitor »

gbnwl wrote: post #10
Wowfunhappy wrote: post #7

The readme says: > This isn't just a tech demo — it's a tool for N64 homebrew developers. Running an LLM natively on N64 hardware enables game mechanics that were impossible in the cartridge era: > AI analyzes play style and adjusts on the fly > NPCs that remember previous conversations and reference past events > In-game level editors where you describe what you want to build ...anyone who has ever used very small l…

You mean to tell me the included screenshot hasn't convinced you? https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...

Delivered. please reconsider now. AI slop cannot build this without a human who has real risc cpu knowledge. The Emulator ---------------------------------------------- https://bottube.ai/watch/shFVLBT0kHY

The real iron! it runs faster on real iron! ---------------------------------------------- https://bottube.ai/watch/7GL90ftLqvh

The Rom image ---------------------------------------------- https://github.com/sophiaeagent-beep/n64llm-legend-of-Elya/b...

reply

Post reply on HN