openmoss/moss-vl

How openmoss/moss-vl works

MOSS-VL is an open-weight video-language model project, with workflows for offline and real-time video inference, supervised fine-tuning, and checkpoint quantization. Users provide video, images, prompts, or training data; inference scripts prepare media and run the model, while training and quantization scripts produce adapted or smaller checkpoints. The graph includes the shared device abstraction and the documented FP8, NF4, and KV-cache quantization branches. Source excerpts are partial, so only relationships shown in excerpts, the source index, or README documentation are drawn.

Main entry points: run_online_inference.py Realtime inference · train.py SFT trainer · run_inference.py Offline inference · quantize_fp8_dynamic.py FP8 quantizer

Components 13

Inference

Offline inference inference/offline/run_inference.py
Realtime inference inference/realtime/run_online_inference.py
Video sources inference/realtime/video_sources.py

Fine-tuning

Training options finetune/arguments.py
SFT dataset finetune/data.py
SFT trainer finetune/train.py

Quantization

FP8 quantizer quantization/quantize_fp8_dynamic.py
NF4 quantizer quantization/quantize_nf4_keep_ends.py
KV8 packager quantization/package_kv8_hqq.py

Runtime support

Device utilities device_utils.py

Other components

User
MOSS-VL checkpoints
Quantized checkpoint

Connections 17

User → Offline inference: runs offline README.md
User → Realtime inference: starts streaming README.md
User → SFT trainer: fine-tunes README.md
Offline inference → MOSS-VL checkpoints: runs inference
Realtime inference → MOSS-VL checkpoints: runs inference
Realtime inference → Video sources: builds frames inference/realtime/run_online_inference.py
SFT trainer → Training options: uses options finetune/train.py
SFT trainer → SFT dataset: loads dataset finetune/train.py
SFT trainer → MOSS-VL checkpoints: fine-tunes checkpoint
User → FP8 quantizer: quantizes checkpoint README.md
User → NF4 quantizer: quantizes checkpoint README.md
User → KV8 packager: packages KV cache README.md
FP8 quantizer → Quantized checkpoint: writes checkpoint
NF4 quantizer → Quantized checkpoint: writes checkpoint
KV8 packager → Quantized checkpoint: packages checkpoint
Offline inference → Device utilities: uses device helpers inference/offline/run_inference.py
Realtime inference → Device utilities: uses device helpers inference/realtime/run_online_inference.py

More C++ diagrams: packtpublishing/mastering-graphics-programming-with-vulkan · universalrobots/universal_robots_ros2_driver · doubangotelecom/ultimatealpr-sdk · diligentgraphics/diligentcore · eldewrito/eldorito

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·