GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
mbn312
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
mbn312
Model Tree
opendatalab/MinerU2.5-Pro-2605-1.2B
Input Modalities
Output Modalities
Supported Functionality
Use the yeasah vLLM EXL3 plugin
with its vLLM fork,
following the plugin's installation instructions. Install mineru-vl-utils==2.0.7
and Pillow for document extraction. Run from the downloaded model directory:
python
import osos.environ["EXL3_INT8_GEMV"] = "1"from PIL import Imagefrom vllm import LLMfrom mineru_vl_utils import MinerUClient, MinerULogitsProcessorfrom mineru_vl_utils.post_process import json2md llm = LLM( model=".", dtype="bfloat16", mm_device_do_normalize=False, logits_processors=[MinerULogitsProcessor],)client = MinerUClient( backend="vllm-engine", vllm_llm=llm, image_analysis=True,)blocks = client.two_step_extract(Image.open("document.png").convert("RGB"))print(json2md([dict(block) for block in blocks]))mm_device_do_normalize=False uses the source image processor's CPU normalization.
The engine uses the packed tied input/output matrix for embedding lookup and generation.
See the original model card for input formats, intended uses and limitations.