Create README.md

README.md

@@ -0,0 +1,22 @@
+- Input: MFCC 13ch, length 100 → shape (B, 13, 100)
+- Delta: (X - mean) / (std + 1e-8)
+- Labels: see `labels.json` (index ↔ label 1:1)
+
+## Usage
+```python
+import json, torch, numpy as np
+from huggingface_hub import hf_hub_download
+from importlib.machinery import SourceFileLoader
+
+repo = "<USER>/audio-emotion-pytorch"
+w = hf_hub_download(repo, "pytorch_model.pth")
+m = hf_hub_download(repo, "model.py")
+lab = hf_hub_download(repo, "labels.json")
+
+labels = json.load(open(lab, encoding="utf-8"))
+Model = SourceFileLoader("amodel", m).load_module().PyTorchAudioModel
+
+model = Model(num_labels=len(labels)).eval()
+state = torch.load(w, map_location="cpu")
+model.load_state_dict(state)
+# x: tensor (1,13,100) → probs = softmax(model(x), dim=1)