1 51 661

Kyle PRO

iky1e

https://ikyle.me

kylehowells

AI & ML interests

None yet

Recent Activity

liked a model about 23 hours ago

VITA-MLLM/VITA-1.5

liked a model about 23 hours ago

VITA-MLLM/Long-VITA-1M

liked a model about 24 hours ago

VITA-MLLM/Long-VITA-128K

View all activity

Organizations

None yet

iky1e's activity

liked 2 models about 23 hours ago

VITA-MLLM/VITA-1.5

Updated 14 days ago • 244 • 21

VITA-MLLM/Long-VITA-1M

Updated 16 days ago • 6

liked a model about 24 hours ago

VITA-MLLM/Long-VITA-128K

Updated 16 days ago • 2

liked a model 1 day ago

PleIAs/pleias_350m_ocr

Updated 1 day ago • 27 • 1

updated a collection 1 day ago

Open Datasets

Collection

33 items • Updated 1 day ago

liked a dataset 1 day ago

cfahlgren1/react-code-instructions

Viewer • Updated 8 minutes ago • 48k • 324 • 78

updated a collection 2 days ago

Text to Speech

Collection

18 items • Updated 2 days ago

liked a model 2 days ago

NeuML/vctk-vits-onnx

Text-to-Speech • Updated Oct 10, 2024 • 268 • 3

reacted to grimjim's post with 🔥 3 days ago

Post

2551

I'm (finally) releasing a Python script that trims excess weights in Gemma2 full-weight models that bloated by ~1B parameters due to an early mergekit bug.
https://github.com/jim-plus/Gemma2-mergekit-remediation

I'd noticed something was off when merges of Gemma2 9B models ended up having ~10B parameters. The current mergekit package is fine, but there are still bloated models on HF that could stand to be fixed.

The script assumes that it will be run from the same directory as the model weights, and will trim the unnecessary lm_head.weight tensor and corresponding index entry.