LoreonLabsPlatform

Overview

Intelligence

Markets
Builders
Research
Ecosystems
Launchpads

Search

Go

vexel

Go library for hosting models on Macs

GoEmerging

Stars

—

Forks

—

Contributors

2

Last push

2mo ago

Recent commits

Latest commits.

docs: add comprehensive LoRA training experiments report
17e000cKeith Ballinger2mo ago
feat(lora): add K/O target modules, data augmentation, and tool-use scenario
e94e419Keith Ballinger2mo ago
test(lora/train): complete 3-scenario eval on Mistral 7B with momentum
7ca6b68Keith Ballinger2mo ago
perf(lora/train): GPU attention weight computation for backward pass
02f99e2Keith Ballinger3mo ago
feat(lora/train): add SGD momentum + fix train/inference chat template mismatch
2ccae1fKeith Ballinger3mo ago

fix(lora/train): fix training OOM on 2B+ models by using pool-based activation buffers

8489120Keith Ballinger3mo ago

test(lora/train): add comprehensive backward pass gradient checking

c35f09bKeith Ballinger3mo ago

fix(lora/train): fix SDPA backward GQA buffer overrun + enable backward pass

9764309Keith Ballinger3mo ago

Top contributors

Builders behind this project.

dependabot[bot]