Any-to-Any
Transformers
Safetensors
qwen3_5
image-text-to-text
video-editing
agent
tool-use
long-horizon
reinforcement-learning
qwen3.5
Instructions to use tencent/WeVisCut with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use tencent/WeVisCut with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("tencent/WeVisCut") model = AutoModelForMultimodalLM.from_pretrained("tencent/WeVisCut", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Update README.md
Browse files
README.md
CHANGED
|
@@ -54,9 +54,6 @@ Training proceeds through unified **SFT**, runtime-verified **RFT**, **Guided Tu
|
|
| 54 |
|
| 55 |
*WeDirector reframes long-form video production as a stateful tool-execution loop — plan, retrieve, compose, inspect, repair — with a shared runtime ledger and multi-round conversational refinement.*
|
| 56 |
|
| 57 |
-
| Checkpoint | Description | Backbone |
|
| 58 |
-
| --- | --- | --- |
|
| 59 |
-
| **WeDirector** (this repo) | Merged RCF-DAPO policy (`unfreeze2`, step 200) for high-quality long-horizon production | Qwen3.5-9B |
|
| 60 |
|
| 61 |
## Performance
|
| 62 |
|
|
|
|
| 54 |
|
| 55 |
*WeDirector reframes long-form video production as a stateful tool-execution loop — plan, retrieve, compose, inspect, repair — with a shared runtime ledger and multi-round conversational refinement.*
|
| 56 |
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
## Performance
|
| 59 |
|