Revert for jart’s llama.cpp MMAP miracles (github.com)

Screenshot of github.com: Revert for jart’s llama.cpp MMAP miracles

First screen, captured 3 years ago.

Looks similar

Screenshot of github.com: Llama.cpp Now Supports Qwen2-VL (Vision Language Model)

Llama.cpp Now Supports Qwen2-VL (Vision Language Model)(github.com)

143 points by BUFU 1 year ago | 37 comments

Screenshot of github.com: New exponent functions that make SiLU and SoftMax 2x faster, at full accuracy

New exponent functions that make SiLU and SoftMax 2x faster, at full accuracy (github.com)

368 points by weinzierl 2 years ago | 72 comments

GitHub pull request to llama.cpp by jart that upstreams llamafile's vectorized expf() functions to compute SiLU and softmax faster and more accurately on CPU.

llama-cpp ggml simd softmax

Screenshot of github.com: Mistral Integration Improved in Llama.cpp

Mistral Integration Improved in Llama.cpp (github.com)

82 points by decide1000 1 year ago | 13 comments

Screenshot of github.com: OpenAI reduces Codex Model Context Size from 372k to 272k

OpenAI reduces Codex Model Context Size from 372k to 272k (github.com)

323 points by AmazingTurtle 3 months ago | 152 comments

A merged GitHub pull request in openai/codex backporting refreshed bundled model metadata (models.json) to the 0.144 release branch.

openai-codex openai context-window llm

Screenshot of github.com: AI has a deep understanding of how this code works

AI has a deep understanding of how this code works (github.com)

212 points by theresistor 11 months ago | 93 comments

Screenshot of github.com: M2 Ultra can run 128 streams of Llama 2 7B in parallel

M2 Ultra can run 128 streams of Llama 2 7B in parallel (github.com)

194 points by behnamoh 3 years ago | 126 comments

Looks similar

Llama.cpp Now Supports Qwen2-VL (Vision Language Model)(github.com)

New exponent functions that make SiLU and SoftMax 2x faster, at full accuracy(github.com)

Mistral Integration Improved in Llama.cpp(github.com)

OpenAI reduces Codex Model Context Size from 372k to 272k(github.com)

AI has a deep understanding of how this code works(github.com)

M2 Ultra can run 128 streams of Llama 2 7B in parallel(github.com)

New exponent functions that make SiLU and SoftMax 2x faster, at full accuracy (github.com)

Mistral Integration Improved in Llama.cpp (github.com)

OpenAI reduces Codex Model Context Size from 372k to 272k (github.com)

AI has a deep understanding of how this code works (github.com)

M2 Ultra can run 128 streams of Llama 2 7B in parallel (github.com)