git
ask
hub
Privacy
Terms
Sign in with GitHub
juncheng0178-del/gwen
↗
Stars ·
30
Language ·
Python
Simplify and Visualize This Repo
Scan the safety of this repo
Self Host this repo
Check out more work of Developer: juncheng0178-del
Ask anything about this repo to start.
I want to fine-tune the GWen 118M model from ModelScope on my own Chinese text dataset. Walk me through the SFT training script and what data format it expects.
Explain how GWen's hybrid attention design works: why does it alternate three Gated DeltaNet layers with one full attention layer every four layers?
I want to run LoRA fine-tuning on GWen to adapt it to a specific domain. Show me how to configure the LoRA training script and which parameters to adjust.
Help me understand GWen's vocabulary design: why use only 8,192 tokens for a Chinese model instead of the typical 50,000+, and what is the tradeoff?
I cloned the GWen repo. Walk me through downloading the pre-trained weights from ModelScope and running the 80M model for Chinese text inference.
Full explanation on explaingit →
📎
Send
By chatting or signing in you agree to the
Terms
and chat-message logging (revocable in
History
).