niansa/libjustlm
1
0
Fork 0
mirror of https://gitlab.com/niansa/libjustlm.git synced 2025-03-06 20:49:17 +01:00
Code Releases Activity
Super easy to use library for doing LLaMA/GPT-J stuff!
93 commits 1 branch 6 tags 544 KiB
C++ 87.3%
CMake 12.5%
C 0.2%
Find a file
Exact
niansa 5b01daa764 Properly implemented MPT
2023-05-15 14:46:19 +02:00
gptj Properly implemented MPT 2023-05-15 14:46:19 +02:00
include Moved gptj/utils.* to g4a-common.* 2023-05-15 09:44:04 +02:00
llama.cpp@0e018fe008 Added GPT-J serialization/deserialization 2023-05-07 12:02:04 +02:00
llama.cpp-alibi@03ceb39c1e Properly implemented MPT 2023-05-15 14:46:19 +02:00
mpt Properly implemented MPT 2023-05-15 14:46:19 +02:00
.gitignore Initial commit 2023-03-30 07:03:33 -05:00
.gitmodules Properly implemented MPT 2023-05-15 14:46:19 +02:00
CMakeLists.txt Properly implemented MPT 2023-05-15 14:46:19 +02:00
g4a-common.cpp Moved gptj/utils.* to g4a-common.* 2023-05-15 09:44:04 +02:00
g4a-common.hpp Moved gptj/utils.* to g4a-common.* 2023-05-15 09:44:04 +02:00
justlm.cpp Properly implemented MPT 2023-05-15 14:46:19 +02:00
justlm_gptj.hpp Moved gptj/utils.* to g4a-common.* 2023-05-15 09:44:04 +02:00
justlm_llama.hpp Check for errors during llama evaluation properly 2023-05-11 18:46:41 +02:00
justlm_mpt.hpp Initial MPT support (buggy) 2023-05-15 12:55:49 +02:00
justlm_pool.cpp Added missing co_await 2023-05-05 00:28:04 +02:00
LICENSE Add LICENSE 2023-04-28 16:08:19 +00:00
pybind.cpp Added GPT-J serialization/deserialization 2023-05-07 12:02:04 +02:00
README.md Add README.md 2023-04-28 16:16:41 +00:00

README.md

JustLM

Super easy to use library for doing LLaMA/GPT-J stuff!

Overview

This library implements an easy to use interface to both LLaMa and GPT-J, with optional Python bindings.

Context scrolling is automatic and supports a top window bar.

Additionally, "pooling" is implemented to support keeping x inference instances in RAM and automatically moving least recently used ones to disk, ready for retrieval.

Documentation

Literally, just read the 2 header files in include/! The interface couldn't be simpler.

Credits

Thanks to Georgi Gerganov (ggerganov) for having written ggml and llama.cpp C libraries, which are both extremely important parts of this project! Also thanks to Nomic AI for having heavily helped me drive this project forward.