Downloads · 30 days
7
12% of all-time downloads
Chrisyichuan/qwen3vl-4b-wiki-screenshot-multik-4x-lora
qwen3vl-4b-wiki-screenshot-multik-4x-lora is a image-text-to-text model from Chrisyichuan. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.
LoRA adapter for Qwen/Qwen3-VL-4B-Instruct, fine-tuned as the reader in a retrieval-augmented Wikipedia-screenshot QA pipeline.
Downloads · 30 days
7
12% of all-time downloads
All-time downloads
58
Public
Repo size
1.3 GB
Likes
0
Public
Click a slice to open those files.
.safetensors1.3 GB · 99%
From the Hugging Face model README
LoRA adapter for Qwen/Qwen3-VL-4B-Instruct, fine-tuned as the reader in a retrieval-augmented Wikipedia-screenshot QA pipeline.
Unlike the earlier fixed-3-image adapters, this v2 reads any number of screenshots from 1 to 6 at inference time, at a fixed 4x pixel compression per image.
| k (images per sample) | base Qwen3-VL-4B @ 0x (no compression, no SFT) | This adapter @ 4x | Δ |
|---|---|---|---|
| 1 | 0.958 | 0.834 | -0.124 |
| 2 | 0.912 | 0.828 | -0.084 |
| 3 | 0.892 | 0.814 | -0.078 |
| 4 | 0.856 | 0.810 | -0.046 |
Base@0x degrades sharply as the number of images grows (distractor confusion: 0.958 → 0.856 from k=1 to k=4). The 4x adapter is nearly k-invariant: 0.834 → 0.810 from k=1 to k=4.
At 4x compression, the per-image pixel budget is ~¼ the original. Even with variable-k SFT, this adapter trails the 2x and 3x variants. Use only under a hard pixel budget.
freeze_vision_tower=false)Qwen/Qwen3-VL-4B-Instructcutoff_len: 6144 (scaled per compression — larger for smaller compression where each image takes more tokens)Built from the Chrisyichuan/screenshot-training-natural-filtered-v2 QA dataset (~104k train examples):
1/sqrt(4) via PIL LANCZOS).<image>×k \n {query} → gold answer.Distribution: ~17k train samples per k value (uniform). eval/test sets likewise stratified over k∈{1..6}.
Gold-retrieval rate at top-6 across splits: ~75%. When gold is present at rank 1..6 among the retrieved hits, composition uses only retrieval results (else gold is always still included by construction).
from transformers import AutoProcessor, Qwen3VLForConditionalGeneration
from peft import PeftModel
import torch
base = "Qwen/Qwen3-VL-4B-Instruct"
adapter = "Chrisyichuan/qwen3vl-4b-wiki-screenshot-multik-4x-lora"
model = Qwen3VLForConditionalGeneration.from_pretrained(base, torch_dtype=torch.bfloat16).cuda()
model = PeftModel.from_pretrained(model, adapter).merge_and_unload()
processor = AutoProcessor.from_pretrained(base)
# k can be any integer in 1..6. Images must already be 4x-compressed.
messages = [{"role": "user", "content": [
{"type": "image", "image": img_1},
# ... up to img_6 ...
{"type": "text", "text": your_question},
]}]
# ... standard Qwen3-VL inference
Chrisyichuan/qwen3vl-4b-wiki-screenshot-multik-2x-lora, Chrisyichuan/qwen3vl-4b-wiki-screenshot-multik-3x-lora.