Skip to content
Other

Qwen3.5-9B Redaktionsfeedback LoRA

Automated Chart Review

2 min readupdated

This is a LoRA (a lightweight adapter built on top of a base model) for the Qwen3.5-9B language model, fine-tuned to review and critique visual data and infographics. Given images of charts—such as line, pie, or bar charts—it analyzes them for errors ranging from missing sources and titles to improper data sorting and color palette issues. The model outputs ready-to-use written feedback with either critiques or quality approval. It is useful for developers and editorial teams looking to automate initial quality checks on graphics before publication.

Key features

  • Built on the modern Qwen3.5-9B base model using the peft library.
  • Trained on a specialized dataset of editorial reviews for charts and diagrams.
  • Detects visualization flaws, including missing legends, swapped axes, and unreadable colors.
  • Generates detailed natural-language feedback with specific notes and suggestions.

What you can do

  • Integrate it into a content pipeline to automatically review newly created charts and diagrams.
  • Build an editorial bot for internal visual QA in Telegram or a web interface.
  • Use it to train designers and analysts on common chart-building mistakes.

Example

An example prompt or infographic workflow: the model receives an image of a pie chart and returns: «Zur Grafik habe ich ein paar Anmerkungen. Es fehlt die Quellenangabe. Bitte unter der Grafik die Quelle nennen (Our World in Data).»

Read moreDetails
Benchmarks and training: best Eval-Loss reached 0.1130 at step 500, with a final Token-Accuracy of 0.9631.
Hyperparameters: Learning Rate 0.0002 (cosine), effective batch size 16, LoRA r=32 / alpha=64 / dropout=0.05 in bf16 format.
Training took roughly 6.19 hours on a dataset of over 10,000 examples (train/val/test).
Examples
Discuss Telegram 0 19

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 18 GB · ready files21 GB
  • FP8 / INT8recommendedweights 9.3 GB11 GB
  • 4-bitweights 5.4 GB6.4 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2080 Ti · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

This is a LoRA: counted together with its base model Qwen3.5-9B.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # JGame3/qwen3.5-9b-lora-20261117-0015
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Qwen/Qwen3.5-9B --local-dir ~/models/Qwen3.5-9B
    hf download JGame3/qwen3.5-9b-lora-20261117-0015 --local-dir ~/models/qwen3.5-9b-lora-20261117-0015
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Qwen3.5-9B
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

Requires the base model

Qwen/Qwen3.5-9B

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. Qwen3.5-9B-Basefamily root
  2. Qwen3.5-9B
  3. qwen3.5-9b-lora-20261117-0015

New posts about versions of Qwen3.5-9B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 6All on Hugging Face
Merges 1All on Hugging Face

About the model

JGame3/qwen3.5-9b-lora-20261117-0015
Author
JGame3
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

190

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Computer vision

VisAlloc-9B

VisAlloc-9B