Code datasets for pretraining
Orion
Orion-zhen
AI & ML interests
Eco-friendly training using Tesla P4. Prefers (FSDP+)QLoRA.
Recent Activity
new activity
6 days ago
grimjim/gemma-3-12b-it-norm-preserved-biprojected-abliterated:Implementation Methodology
updated
a Space
about 1 month ago
Orion-zhen/mcpo
updated
a Space
about 2 months ago
Orion-zhen/webgpuf
Organizations
AWQ quantization of Qwen3 Dense series at Day0!
All for emoji!
Datasets used for various calibrations
Datasets used to unalign models
Powerful apps built on free HF space
My llama3 models
-
Orion-zhen/Llama3-70B-Orion-Chinese
Text Generation • 71B • Updated • 51 • • 14 -
Orion-zhen/Llama3-70B-Orion-Chinese-SE
Text Generation • 71B • Updated • 13 -
Orion-zhen/Llama3-70B-Orion-Chinese-Plus
Text Generation • 71B • Updated • 9 -
Orion-zhen/Llama3-70B-Orion-Chinese-Ultra
Text Generation • 71B • Updated • 7 • 1
Datasets focus on reasoning
Code datasets for pretraining
Powerful apps built on free HF space
AWQ quantization of Qwen3 Dense series at Day0!
All for emoji!
My llama3 models
-
Orion-zhen/Llama3-70B-Orion-Chinese
Text Generation • 71B • Updated • 51 • • 14 -
Orion-zhen/Llama3-70B-Orion-Chinese-SE
Text Generation • 71B • Updated • 13 -
Orion-zhen/Llama3-70B-Orion-Chinese-Plus
Text Generation • 71B • Updated • 9 -
Orion-zhen/Llama3-70B-Orion-Chinese-Ultra
Text Generation • 71B • Updated • 7 • 1
Datasets used for various calibrations
Datasets focus on reasoning
Datasets used to unalign models