Upload folder using huggingface_hub

README.md

@@ -12,28 +12,55 @@ pipeline_tag: other
 
 # Sonny
 
-StepNets one-step weather backbone: variable-aware embedding and a two-stage trunk, 69 ERA5 variables, 120×240, patch size 2. 
+StepNets one-step weather backbone: variable-aware embedding and a two-stage trunk, 69 ERA5-like variables, 120×240, patch size 2. EMA checkpoint at epoch 50.
 
-## Loading (with the `stormer` codebase)
+## Loading
 
-This Hub snapshot stores weights only (`model.safetensors`) plus `config.json` (`init_args`). 
-You need the same Python dependencies as training (**PyTorch**, **timm**, **xformers** on GPU, etc.) and the `stormer` package that defines the model.
+Weights: `model.safetensors`. Hyperparameters: `config.json` (`init_args`).  
+Install the small **`sonny`** package (model code only; [GitHub](https://github.com/YOUR_ACCOUNT/sonny)) or the full `stormer` research tree if you already use it.
 
 ```python
-from stormer.models.hub.sonny import load_sonny_from_hub
+from sonny import load_sonny_from_hub
 
 model = load_sonny_from_hub("Francismj/sonny", map_location="cuda")
 model.eval()
 ```
 
-Or download files manually and build :class:`~stormer.models.hub.sonny.Sonny`(**init_args**) then :meth:`load_state_dict`.
+Or download the files and build `networks.sonny.Sonny(**init_args)` then `load_state_dict` (see `class_qualname` in `config.json`).
+
+## Google Colab
+
+Install `sonny` (clone or zip of the repo with `pyproject.toml`, packages `sonny/` and `networks/`).
+
+```text
+!pip install -q huggingface_hub safetensors timm pyyaml
+!pip install -q xformers
+!git clone https://github.com/YOUR_ACCOUNT/sonny.git sonny_src
+%cd sonny_src
+!pip install -q -e .
+%cd /content
+```
+
+```python
+from sonny import load_sonny_from_hub
+
+model = load_sonny_from_hub("Francismj/sonny", map_location="cuda")
+model.eval()
+```
+
+Private code: zip the `sonny` repo → upload in Colab → `pip install -e /content/unzipped_folder`.
 
 ## Files
 
-- `config.json` — hyperparameters / variable list for **Sonny**
+- `config.json` — hyperparameters and variable list
 - `model.safetensors` — EMA ``ema_model_state_dict``
 - `normalization_note.txt` — reminder to use your local ``normalization_constants`` from the training repo for real inference
 
 ## Citation
 
-If you use this checkpoint, cite [Sonny](https://arxiv.org/abs/2603.21284).
+If you use this checkpoint or architecture, cite:
+
+**Sonny: Breaking the Compute Wall in Medium-Range Weather Forecasting** — Minjong Cheon, arXiv:2603.21284, 2026.
+[https://arxiv.org/abs/2603.21284](https://arxiv.org/abs/2603.21284)
+
+Code: [https://github.com/jmj2316/sonny](https://github.com/jmj2316/sonny)

config.json

@@ -3,7 +3,7 @@
   "sonny_version": "1.0",
   "architecture_name": "Sonny",
   "architecture_family": "StepNets",
-  "class_qualname": "stormer.models.hub.sonny.Sonny",
+  "class_qualname": "networks.sonny.Sonny",
   "source_checkpoint": "stormbreaker5_checkpoint_ema_epoch_050.pth",
   "source_epoch": 50,
   "init_args": {