BeFM / app.py

Commit History

Update default temperature from 0.6 to 0.7
49689a5

Jn-Huang Claude Sonnet 4.5 commited on

Add custom chatbot component to chat interface
3db6d3c

Jn-Huang Claude Sonnet 4.5 commited on

Update prompt reference text and space emoji
97c12a9

Jn-Huang commited on

Align placeholder with default system prompt
d8ad02c

Jn-Huang commited on

Update default prompt text and appendix link
1002aec

Jn-Huang commited on

Update default prompt and fix chat interface inputs
d298fc0

Jn-Huang commited on

update messages
77a4a60

Jn-Huang commited on

Customize BeFM UI and defaults
58ffc70

Jn-Huang commited on

Make code compatible with Gradio 6.0 structured content format
38dedc7

Jn-Huang commited on

Fix Gradio ChatInterface: remove lambda wrapper, add lazy loading, make public
4cc1531

Jn-Huang commited on

Switch to transformers version - vLLM uses too much memory on T4 GPU
fc3b3a2

Jn-Huang commited on

Reduce vLLM GPU memory utilization to 0.7 to avoid OOM on T4 GPU
0600d50

Jn-Huang commited on

Switch to vLLM for faster inference with lazy loading and multi-turn fix
89babab

Jn-Huang commited on

Fix multi-turn conversation: handle dict history format
eaaeae1

Jn-Huang commited on

Fix bugs: use token param, apply Llama 3.1 chat template, decode only new tokens
1a77428

Jn-Huang commited on

Add Be.FM-8B chat interface with PEFT adapter
f6fde6f

Jn-Huang commited on