Alan Tseng's picture

2 4

Alan Tseng

agentlans

·

agentlans

AI & ML interests

Small data, boring AI

Recent Activity

updated a model 24 minutes ago

agentlans/Gemma2-9B-AdvancedFuse

reacted to grimjim's post with 👍 about 3 hours ago

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.

replied to grimjim's post about 3 hours ago

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.

View all activity

Organizations

None yet

agentlans's activity

New activity in agentlans/Qwen2.5-0.5B-Instruct-CrashCourse-dropout 2 days ago

Adding Evaluation Results

#1 opened 2 days ago by

New activity in agentlans/Llama3.1-8B-drill 6 days ago

Adding Evaluation Results

#1 opened 6 days ago by

Adding Evaluation Results

#1 opened 6 days ago by