Downloads · 30 days
13
30% of all-time downloads
OnAnOrange/CUA-Lite-Qwen3.5-4B-ScaleCUA-gpt5_5-epoch1-public
CUA-Lite-Qwen3.5-4B-ScaleCUA-gpt5_5-epoch1-public is a image-text-to-text model from OnAnOrange. 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.
Full-parameter SFT checkpoint of Qwen3.5-4B, trained with CUA-Lite on 5,000 Lite.ScaleCUA desktop trajectories from teacher gpt55. This repository contains the checkpoint after epoch 1 of a three-epoch training run, w…
Downloads · 30 days
13
30% of all-time downloads
All-time downloads
44
Public
Parameters
4.7B
9.3 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors9.3 GB · 100%
From the Hugging Face model README
Full-parameter SFT checkpoint of Qwen3.5-4B, trained with CUA-Lite on 5,000 Lite.ScaleCUA desktop trajectories from teacher gpt5_5. This repository contains the checkpoint after epoch 1 of a three-epoch training run, with the full BF16 model, tokenizer, and image processor configuration.
| Setting | Value |
|---|---|
| Teacher / dataset variant | desktop.use.train.gpt5_5 |
| Training trajectories | 5,000 |
| Serialized trajectory steps | 42,427 |
| Completed epochs | 1 |
| Completed optimizer updates | 156 |
| Source checkpoint | iter_155 |
| Last-step training loss | 0.13092419 |
| Batch size | 32 trajectories; microbatch 1 |
| Hardware and parallelism | 8 H200 GPUs; TP=2, DP=4 |
| Learning rate | 5e-6; cosine decay to 1e-6; 10% warmup |
| Optimizer | Adam, betas=(0.9, 0.95), weight decay=0.1 |
| Precision | BF16 |
| Training date | 2026-09-09 |
The action-only desktop recipe uses a 1280x720 screenshot target (rounded to 1280x704 by the processor), history length 50, and up to four prompt images per step. Reasoning is disabled. Data is filtered for non-excluded trajectories with episode return greater than 0.5. Conversion failures are skipped and the seeded candidate pool is replenished until 5,000 usable trajectories are obtained.
The trainer defines an epoch as floor(5000 / 32) = 156 optimizer updates. One checkpoint is saved per epoch, at zero-based rollout IDs 155, 311, and 467. The last-step training loss above is a single training-batch value, not a benchmark score. No downstream benchmark evaluation is reported for this checkpoint.
851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a47deef05c5525d98fcb79399c6c33215925e4479Use a Transformers version with Qwen3.5 support; the training export used Transformers 5.6.0. Load the model and processor from this repository. For desktop-agent inference, use CUA-Lite with the included desktop.use.compact.yaml so the prompt, action format, and image history match training. Full training details are in training_config.json.
The original Qwen3.5-4B Apache 2.0 license is included as LICENSE. These weights are modified by supervised fine-tuning as described above.