Upload 3 files

config.json

@@ -0,0 +1,51 @@
+{
+  "model_type": "small_audio_classifier_mlp",
+  "torch_dtype": "float32",
+  "input_dim": 6144,
+  "num_labels": 10,
+  "hidden_dims": [
+    512,
+    256
+  ],
+  "dropout": 0.2,
+  "activation": "gelu",
+  "task": "audio_classification",
+  "outputs": {
+    "logits": "float32[batch, num_labels]"
+  },
+  "preprocessing": {
+    "expected_input": "flattened log-mel spectrogram or other fixed-length audio features",
+    "example_logmel": {
+      "n_mels": 64,
+      "n_frames": 96,
+      "flatten": true,
+      "input_dim": 6144
+    },
+    "note": "compute features outside the model; this model is intentionally tiny for assignments"
+  },
+  "id2label": {
+    "0": "class_0",
+    "1": "class_1",
+    "2": "class_2",
+    "3": "class_3",
+    "4": "class_4",
+    "5": "class_5",
+    "6": "class_6",
+    "7": "class_7",
+    "8": "class_8",
+    "9": "class_9"
+  },
+  "label2id": {
+    "class_0": 0,
+    "class_1": 1,
+    "class_2": 2,
+    "class_3": 3,
+    "class_4": 4,
+    "class_5": 5,
+    "class_6": 6,
+    "class_7": 7,
+    "class_8": 8,
+    "class_9": 9
+  },
+  "version": "0.1.0"
+}

model.py

@@ -0,0 +1,38 @@
+import json
+import torch
+import torch.nn as nn
+
+class SmallAudioClassifierMLP(nn.Module):
+    def __init__(self, input_dim=6144, num_labels=10, hidden_dims=(512, 256), dropout=0.2, activation="gelu"):
+        super().__init__()
+        act = {"relu": nn.ReLU, "gelu": nn.GELU, "silu": nn.SiLU}[activation]
+        dims = [input_dim] + list(hidden_dims)
+        layers = []
+        for i in range(len(dims) - 1):
+            layers += [
+                nn.Linear(dims[i], dims[i+1]),
+                nn.LayerNorm(dims[i+1]),
+                act(),
+                nn.Dropout(dropout),
+            ]
+        self.mlp = nn.Sequential(*layers)
+        self.classifier = nn.Linear(dims[-1], num_labels)
+
+    def forward(self, x):
+        h = self.mlp(x)
+        return self.classifier(h)
+
+def load_pretrained(model_dir: str, map_location="cpu"):
+    with open(f"{model_dir}/config.json", "r", encoding="utf-8") as f:
+        cfg = json.load(f)
+    m = SmallAudioClassifierMLP(
+        input_dim=cfg["input_dim"],
+        num_labels=cfg["num_labels"],
+        hidden_dims=tuple(cfg["hidden_dims"]),
+        dropout=cfg["dropout"],
+        activation=cfg["activation"],
+    )
+    sd = torch.load(f"{model_dir}/pytorch_model.bin", map_location=map_location)
+    m.load_state_dict(sd)
+    m.eval()
+    return m, cfg

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d2758b33b878bcca1883b88f69b5dcce7f905a8dd5ed17df8a093af16a36996
+size 13130572