Downloads · 30 days
7
15% of all-time downloads
c43g-s600m2
sjin59/c43g-s600m2
c43g-s600m2 is a image-text-to-text model from sjin59. 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 transformers. The card lists the license as apache-2.0.
Given 4 shuffled frames from a video plus a caption describing what happens, this model recovers the original chronological order of the frames.
All-time downloads
47
Public
Parameters
27.4B
54.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors54.7 GB · 100%
At a glance
- Task
- Image-Text-to-Text
- Library
- transformers
- License
- apache-2.0
- Model type
- qwen3_5
- Access
- Public
- Created
- Jul 27, 2026
- Updated
- Jul 27, 2026
- SHA
- d1edfdc7
Try a prompt
Base models
Model card
From the Hugging Face model README
c43g-s600m2
Given 4 shuffled frames from a video plus a caption describing what happens, this model recovers the original chronological order of the frames.
Fine-tuned from Qwen3.5-27B (VLM):
Qwen3.5-27B
→ non-thinking full SFT (16,070 rows, seed 43, 1,004 steps)
→ non-thinking GRPO LoRA (r32/α64, 1,400 steps; checkpoint-600 selected)
→ LoRA merge = this model
Output is four letters in chronological order, e.g. "C,B,A,D".
Usage
Runs on vLLM 0.19.0. In bf16 it needs 2× H100 80GB (TP2). For single-GPU inference use the GGUF build, which fits on one RTX 3090 24GB.
Default sampling is temperature 0.6 / top_p 0.95 / top_k 20 — not greedy decoding.
Licensed under Apache-2.0, following the base model Qwen3.5-27B.
- Task
- Image-Text-to-Text
- Library
- transformers
- Type
- qwen3_5
- License
- apache-2.0
- Languages
- en
- Created
- Jul 27, 2026
- Updated
- Jul 27, 2026