Infinity-Parser2-Pro — AWQ W4A16 (int4)
An int4 (W4A16) AWQ quantization of infly/Infinity-Parser2-Pro, made to run on NVIDIA A100 / sm80, where the original FP8 path is unsupported. ~70 GB bf16 → ~21 GB.
Method
AWQ via llm-compressor, routed experts only (W4A16). Attention, GDN/linear-attention, shared expert, vision tower, and lm_head are kept in bf16. Calibrated on a few hundred diverse document + general-vision samples.
Serving note (important)
vLLM fuses some bf16 layers before consulting the ignore list, which can otherwise yield all-! output. Fix: the saved config.json quantization_config.ignore uses broad regexes matching the fused names. Already applied here.
Quality (VLMEvalKit, AI-judged reproduction)
| Benchmark | Published bf16 | This int4 |
|---|---|---|
| MMStar | 69.7 | 66.9 |
| OCRBench | 86.2 | ~89 |
| DocVQA (val) | 96.4 | 96.4 |
Near-lossless on the cleanly-comparable axes. (MMBench omitted to avoid a circular-vs-vanilla scoring mismatch.)
Usage (vLLM)
vllm serve spectator2026/Infinity-Parser2-Pro-AWQ-W4A16 --dtype bfloat16 --trust-remote-code --reasoning-parser qwen3
Pass chat_template_kwargs={"enable_thinking": false} in requests, or answers land in the reasoning channel.
Quantized by @spectator2026. Original model © infly, Apache-2.0.