Darkstar-Qwen-Image-2.1-Base-ModelOpt-FP8

Non-commercial research artifact. NVIDIA ModelOpt FP8 (W8A8, static scales) quantization of Qwen/Qwen-Image-2.1, built by the Model Forge pipeline from the pinned upstream BF16 — no third-party quantized checkpoints were used as build inputs.

Published revision: 5fcc2176bd2edc19d27a33229f75ff2a03decdfb

License — Qwen Research License Agreement (read this first)

This is a derivative of Qwen-Image-2.1 and inherits the Qwen Research License Agreement (September 20, 2026). The full agreement text is shipped in this repository as LICENSE.

Qwen is licensed under the Qwen RESEARCH LICENSE AGREEMENT, Copyright (c) 2026 Hangzhou Tongyi Laboratory Technology Co., Ltd. All Rights Reserved.

Obligation	Meaning for you
§1(i)	Use for research or evaluation purposes only — non-commercial
§3	Retain this Notice in distributions; prominent notices on modified files
§4a	You are responsible for export-control compliance
§4b	Products built with this model display "Built with Qwen"; products trained from it display "Improved using Qwen"
§4c	Do not name competing primary models "Qwen"
§5a	Derivatives you create are owned by you, subject to this agreement
§5c	License terminates if you initiate litigation against the licensor

Commercial use requires a separate license from the Qwen team (model-business@notice.qwencloud.com).

Summary

FP8 (W8A8, static weight+activation scales) quantization of the 7B single-stream DiT (QwenImage21Transformer2DModel, 32 blocks) — the CI-fast recipe validating the whole quantize → export → serve → evaluate loop; the NVFP4 cell is the primary release. Text encoder (Qwen3-VL 8B), VAE, norms/embeddings/shared-modulation stay BF16.

Precision map

Component	Precision
DiT blocks 0–31 linear layers	FP8-e4m3 (static weight + input scales)
img_in, txt_in, modulation, time_text_embed, norm_out, proj_out	BF16
Per-tensor scales	fp32 weight_scale + input_scale
Text encoder (Qwen3-VL 8B)	BF16
VAE	BF16

Provenance

• Upstream model: Qwen/Qwen-Image-2.1
• Upstream revision: b3179ad355be050328e483a9dfdd9e60cd62adfa
• Weight edit: ModelOpt FP8 quantization (own pipeline; calibration = Gustavosta Stable-Diffusion-Prompts, max algo)
• Toolchain: nvidia-modelopt 0.46.0rc2 @ 43fd41a… + diffusers main @ 80c7ed26… (bridge-fork overlay registering the 2.1 ModelType; sha-pinned)
• Recipe: darkstar-qwen-image-2-1-base-fp8.yaml (in the engineering repository)
• Engineering repository: HangGlidersRule/model-forge

Serving (vLLM-Omni — validated)

docker run --gpus all -p 8000:8000 \
  -e NCCL_CUMEM_ENABLE=0 \
  -v <hf-cache>:/root/.cache/huggingface \
  vllm/vllm-omni:qwen-image21 \
  bash -c 'vllm serve HangGlidersRule/Darkstar-Qwen-Image-2.1-Base-ModelOpt-FP8 --omni --port 8000'

OpenAI-compatible endpoints: POST /v1/images/generations, POST /v1/images/edits.

Note: the checkpoint config carries component-prefixed ignore patterns (*transformer_blocks.N.*). The vLLM-Omni runtime matches quantization exclusions against component-prefixed module names; unprefixed patterns silently fail to exclude (known upstream issue, documented by Model Forge).

Consumption without a server (diffusers / ComfyUI-style)

diffusers from_pretrained on this repo fails on the current diffusers main ModelOpt-quantizer schema (known upstream gap). Working consumption paths:

1. vLLM-Omni (above) — the validated serving path.
2. Dequant-on-load (documented reference implementation in the Model Forge repo): strip quantization_config from a local copy's transformer/config.json, load the pipeline, then dequantize the fp8 weights (w_fp8 × weight_scale). Validated end-to-end below.

Evaluation (frozen protocol)

Quality manifest image-eval-v1 (sha 09daf654e4015a49d1f54961687d4f6df41da9910e4c0c25d1981582190aeee9), 25 frozen cases, full-denominator scoring:

Capability	BF16 control	This cell
t2i_quality	10/10	10/10
composition	6/6	6/6
typography	4/4	4/4
rgba	2/2	2/2
editing	3/3	3/3

• Quant validators: 196 quantized modules, 0 misclassified vs policy, 0 degenerate scales.
• Dequant math verified exact vs upstream BF16 (max err = FP8 quant noise).
• Serve smoke: healthy → HTTP 200 in 15.5s → visually verified coherent.
• Clean-download boot smoke from this repository: sha-verified + generation PASS.

Family context

The Darkstar Qwen-Image-2.1 family ships a four-cell matrix plus FP8 variants. The NVFP4 cell is the primary release; this FP8 cell is the CI-fast sibling with identical behavior. Ablit cells are not part of this release (DiT-side behavior-transform research in the Model Forge private ledger).

Non-commercial. Research and evaluation use only. Built with Qwen.