Look Ma, No Bubbles: Designing a Low-Latency Megakernel for Llama-1B (hazyresearch.stanford.edu)

Screenshot of hazyresearch.stanford.edu: Look Ma, No Bubbles: Designing a Low-Latency Megakernel for Llama-1B

First screen, captured 1 year ago.

Looks similar

Screenshot of hazyresearch.stanford.edu: GPUs Go Brrr

GPUs Go Brrr (hazyresearch.stanford.edu)

1056 points by nmstoker 2 years ago | 243 comments

Stanford Hazy Research blog post on what GPU hardware wants for efficient AI kernels, introducing the ThunderKittens CUDA kernel library.

gpu-kernels thunderkittens cuda flashattention

Screenshot of hazyresearch.stanford.edu: AMD GPUs Go Brrr

AMD GPUs Go Brrr (hazyresearch.stanford.edu)

222 points by vinhnx 11 months ago | 71 comments

Screenshot of hazyresearch.stanford.edu: We bought the whole GPU, so we're damn well going to use the whole GPU

We bought the whole GPU, so we're damn well going to use the whole GPU (hazyresearch.stanford.edu)

472 points by sydriax 1 year ago | 94 comments

Stanford Hazy Research blog post introducing a throughput-optimized megakernel for tensor-parallel Llama-70B inference on H100 GPUs that overlaps compute, memory, and communication.

megakernels llm-inference tensor-parallelism llama-70b

Screenshot of minimaxir.com: The mysterious Hy3 LLM is topping OpenRouter Model Rankings by a large margin

The mysterious Hy3 LLM is topping OpenRouter Model Rankings by a large margin (minimaxir.com)

131 points by freediver 5 months ago | 103 comments

Screenshot of hazyresearch.stanford.edu: How to scale LLMs better with an alternative to transformers

How to scale LLMs better with an alternative to transformers (hazyresearch.stanford.edu)

142 points by tuxguy 3 years ago | 30 comments

Screenshot of interconnects.ai: Llama 2: an open-source LLM

Llama 2: an open-source LLM (interconnects.ai)

35 points by tim_sw 3 years ago | 4 comments

Looks similar

GPUs Go Brrr(hazyresearch.stanford.edu)

AMD GPUs Go Brrr(hazyresearch.stanford.edu)

We bought the whole GPU, so we're damn well going to use the whole GPU(hazyresearch.stanford.edu)

The mysterious Hy3 LLM is topping OpenRouter Model Rankings by a large margin(minimaxir.com)

How to scale LLMs better with an alternative to transformers(hazyresearch.stanford.edu)

Llama 2: an open-source LLM(interconnects.ai)