calcuis/cow-encoder

🐮cow architecture gguf encoder

• don't need to rebuild tokenizer from metadata
• don't need separate tokenizer file
• no more oom issues (possibly)

eligible model example

• use cow-gemma2 2.33GB for lumina
• use cow-umt5base 451MB for ace-audio
• use cow-umt5xxl 3.67GB for wan-s2v or any wan video model

the example workflow above is from wan-s2v-gguf; cow encoder is a special designed clip, even the lowest q2 quant still working very good; upgrade your node for cow-encoder support🥛🐮 and do drink more milk

Prompt

a conversation between cgg and connector

reference

• gguf-node (pypi|repo|pack)