Skip to content
Computer vision

Lucia 2B v1.2-nc

chain-of-thought image analysis

2 min read

A new LoRA adapter (a small add-on for a base model) for Qwen/Qwen3.5-2B that trains the language model to carefully examine images and output structured breakdowns: step-by-step thoughts, captions, tags, prompts, and Q&As. Version 1.2-nc is a non-commercial counterpart to the Lucia 2B v1.2 model and is trained on an expanded dataset of over 184,000 examples. Thanks to a special training objective, the model preserves base model behavior on standard text queries and only outputs breakdowns when needed. Useful for developers and researchers working on detailed image descriptions and multimodal data tasks.

Key features

  • Trained on 184,562 examples from the thought-symbolic-image-recaptioned dataset.
  • Maintains standard Qwen3.5-2B text-prompt behavior through a special training objective.
  • Supports variable analysis depth: from quick glances to detailed breakdown with critique and questions.
  • Improved text recognition on images compared to the first version.

What you can do

  • Use the adapter in local pipelines to generate detailed prompts, tags, and image descriptions.
  • Automate visual data parsing in research projects (subject to the CC BY-NC-SA 4.0 non-commercial license).
  • Integrate the model with transformers and peft libraries to analyze photo and scene structure.

Example

import torch
from peft import PeftModel
from PIL import Image
from transformers import AutoModelForMultimodalLM, AutoProcessor

processor = AutoProcessor.from_pretrained("AbstractPhil/qwen3.5-2b-think-symbolic-image-nc")
model = AutoModelForMultimodalLM.from_pretrained("Qwen/Qwen3.5-2B", dtype=torch.bfloat16, device_map="cuda")
model = PeftModel.from_pretrained(model, "AbstractPhil/qwen3.5-2b-think-symbolic-image-nc").merge_and_unload().eval()

image = Image.open("picture.jpg").convert("RGB")
prompt = ("Look at this picture and think it through in writing, a closer look; then write a caption and a prompt or tags, "
          "and the questions a careful looker would still ask, each as `question: what -> answer [record, chain or unknown]`.")
messages = [{"role": "user", "content": [{"type": "image", "image": image}, {"type": "text", "text": prompt}]}]
text = processor.apply_chat_template(messages, add_generation_prompt=True, tokenize=False, enable_thinking=False)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=800, do_sample=False)
print(processor.batch_decode(out[:, inputs["input_ids"].shape[1]:], skip_special_tokens=True)[0])
Read moreDetails
Built on Qwen/Qwen3.5-2B with a LoRA adapter trained for 21,815 steps.
Requires transformers 5 and peft 0.18 or newer, plus a CUDA-capable GPU with bfloat16 support.
Distributed under the CC BY-NC-SA 4.0 license (non-commercial use only due to the image dataset).
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 4.4 GB · ready files5.3 GB
  • FP8 / INT8recommendedweights 2.2 GB3.0 GB
  • 4-bitweights 1.3 GB1.9 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 2060 · BF16 — from RTX 2060

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 Qwen3.5-2B.

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.

    # AbstractPhil/qwen3.5-2b-think-symbolic-image-nc
    # ===== 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/Qwen3.5-2B --local-dir ~/models/Qwen3.5-2B
    hf download AbstractPhil/qwen3.5-2b-think-symbolic-image-nc --local-dir ~/models/qwen3.5-2b-think-symbolic-image-nc
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Qwen3.5-2B
  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/Qwen3.5-2B

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

  1. Qwen3.5-2B-Basefamily root
  2. Qwen3.5-2B
  3. qwen3.5-2b-think-symbolic-image-nc

New posts about versions of Qwen3.5-2B-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 10All on Hugging Face
Merges 2All on Hugging Face

About the model

AbstractPhil/qwen3.5-2b-think-symbolic-image-nc
Author
AbstractPhil
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments