Metadata-Version: 2.4
Name: lightvl
Version: 0.1.1
Summary: lightvl is a lightweight Vision-Language Model (VLM) quantization toolkit
Author: alexzhan
Author-email: 
Description-Content-Type: text/markdown
Requires-Dist: vllm>=0.12.0
Requires-Dist: prettytable>=3.12.0
Requires-Dist: transformers>=4.57.1
Dynamic: author
Dynamic: description
Dynamic: description-content-type
Dynamic: requires-dist
Dynamic: summary

LightVL is a lightweight Vision-Language Model (VLM) quantization toolkit supporting FP8, INT8, FP8-Block, and INT4-FP8 mixed quantization schemes. It integrates with vLLM for high-throughput inference and supports Qwen3-VL, Qwen3.5, InternVL-Chat, and Gemma-4 models.

Key capabilities:
- CPU or GPU fast quantization
- Automated side-by-side comparison of original vs quantized model latency and accuracy  based vllm and transformers
- Batch inference with concurrent request execution
