Upload folder using huggingface_hub

config.json

@@ -0,0 +1,28 @@
+{
+  "model_type": "crepe",
+  "architecture": "CREPE-full",
+  "base_model": "marl/crepe (Kim et al. 2018)",
+  "library": "torchcrepe",
+  "library_version": "0.0.24",
+  "input_sample_rate": 16000,
+  "input_window_size": 1024,
+  "num_output_bins": 360,
+  "bin_resolution_cents": 20,
+  "pitch_range_hz": {
+    "min": 32.7,
+    "max": 1975.5
+  },
+  "model_size": "full",
+  "num_parameters": 22244000,
+  "finetuned_from": "torchcrepe pretrained CREPE-full",
+  "finetuning_datasets": [
+    "MOSA",
+    "MusicNet (Option B Solo Violin)"
+  ],
+  "target_domain": "violin (monophonic)",
+  "metrics": {
+    "Bach10_RPA": 0.982,
+    "MOSA_test_RPA": 0.862,
+    "MusicNet_test_RPA": 0.309
+  }
+}

convert_to_hf.py

@@ -0,0 +1,99 @@
+import json
+import torch
+from pathlib import Path
+
+# Load best.pt
+print("Loading best.pt ...")
+state = torch.load("best.pt", map_location="cpu", weights_only=False)
+
+# Extract state_dict
+if isinstance(state, dict):
+    if "model" in state:
+        state_dict = state["model"]
+        print("  found state_dict under 'model'")
+    elif "state_dict" in state:
+        state_dict = state["state_dict"]
+        print("  found state_dict under 'state_dict'")
+    else:
+        state_dict = state
+        print("  treating dict as state_dict")
+else:
+    state_dict = state
+print(f"  {len(state_dict)} tensors")
+
+# Save safetensors
+try:
+    from safetensors.torch import save_file
+except ImportError:
+    import subprocess
+    subprocess.check_call(["pip3", "install", "safetensors"])
+    from safetensors.torch import save_file
+
+state_dict = {k: v.contiguous() for k, v in state_dict.items()}
+save_file(state_dict, "model.safetensors")
+print(f"OK saved model.safetensors")
+
+# config.json
+config = {
+    "model_type": "crepe",
+    "architecture": "CREPE-full",
+    "base_model": "marl/crepe (Kim et al. 2018)",
+    "library": "torchcrepe",
+    "library_version": "0.0.24",
+    "input_sample_rate": 16000,
+    "input_window_size": 1024,
+    "num_output_bins": 360,
+    "bin_resolution_cents": 20,
+    "pitch_range_hz": {"min": 32.7, "max": 1975.5},
+    "model_size": "full",
+    "num_parameters": 22244000,
+    "finetuned_from": "torchcrepe pretrained CREPE-full",
+    "finetuning_datasets": ["MOSA", "MusicNet (Option B Solo Violin)"],
+    "target_domain": "violin (monophonic)",
+    "metrics": {
+        "Bach10_RPA": 0.982,
+        "MOSA_test_RPA": 0.862,
+        "MusicNet_test_RPA": 0.309
+    }
+}
+with open("config.json", "w") as f:
+    json.dump(config, f, indent=2, ensure_ascii=False)
+print("OK saved config.json")
+
+# preprocessor_config.json
+preprocessor = {
+    "feature_extractor_type": "CrepeFrameExtractor",
+    "sampling_rate": 16000,
+    "channels": 1,
+    "window_size_samples": 1024,
+    "hop_size_samples": 160,
+    "hop_size_ms": 10,
+    "window_size_ms": 64,
+    "normalization": "z_score",
+    "normalize_per_frame": True,
+    "epsilon": 1e-6,
+    "expected_format": "wav",
+    "expected_sample_rate": 16000,
+    "expected_channels": "mono",
+    "resample_if_needed": True,
+    "decoder": {
+        "method": "weighted_argmax",
+        "neighborhood_size": 9,
+        "smoothing": {
+            "periodicity_median_filter": 3,
+            "pitch_mean_filter": 3
+        },
+        "voicing_threshold": 0.21
+    }
+}
+with open("preprocessor_config.json", "w") as f:
+    json.dump(preprocessor, f, indent=2, ensure_ascii=False)
+print("OK saved preprocessor_config.json")
+
+print("\nDone! Files created:")
+for f in ["model.safetensors", "config.json", "preprocessor_config.json"]:
+    size = Path(f).stat().st_size
+    if size > 1024 * 1024:
+        print(f"  {f}: {size/1024/1024:.1f} MB")
+    else:
+        print(f"  {f}: {size} bytes")

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5ebae0cd5aeff1735bf248d78fecca99a31a8dc8c6b97c8c0d418a63e3a2534
+size 88981056

preprocessor_config.json

@@ -0,0 +1,25 @@
+{
+  "feature_extractor_type": "CrepeFrameExtractor",
+  "sampling_rate": 16000,
+  "channels": 1,
+  "window_size_samples": 1024,
+  "hop_size_samples": 160,
+  "hop_size_ms": 10,
+  "window_size_ms": 64,
+  "normalization": "z_score",
+  "normalize_per_frame": true,
+  "epsilon": 1e-06,
+  "expected_format": "wav",
+  "expected_sample_rate": 16000,
+  "expected_channels": "mono",
+  "resample_if_needed": true,
+  "decoder": {
+    "method": "weighted_argmax",
+    "neighborhood_size": 9,
+    "smoothing": {
+      "periodicity_median_filter": 3,
+      "pitch_mean_filter": 3
+    },
+    "voicing_threshold": 0.21
+  }
+}