v2.0.1

Disable 8-bit quantization and fix Llama-2 output parsing

v2.0.0

Upgrade to MiniGPT-v2 (Llama-2-7b-chat) for better instruction following

v1.2.5

Disable sampling (do_sample=False) to fix probability tensor error

v1.2.4

Add detailed logging to handler.py to debug crash

v1.2.3

Increase image resolution to 448x448 for better document reading

v1.2.2

Tune generation parameters: use beam search (5), low temp (0.1) to re…

v1.2.1

Switch to RunPod base image and upgrade bitsandbytes

v1.2.0

Add debug logging and fix base64 image handling

v1.1.9

Fix handler.py: manually configure visual processor to bypass missing…

v1.1.8

Switch to devel image and set LD_LIBRARY_PATH for bitsandbytes