Upload folder using huggingface_hub

GridshiftCLAP.mlpackage.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5512b6fa4aab44b2a08dcb505ae31521b20f849025b2c8a941e81fc4e543f21e
+size 66006191

GridshiftCLAPText.mlpackage.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f2d1bababce0e0fc74f6934ea46b5e5dcfd99167213600276f60285ec87257d
+size 112265825

README.md

@@ -0,0 +1,16 @@
+---
+license: apache-2.0
+base_model: laion/larger_clap_general
+tags: [coreml, clap, audio, embeddings]
+---
+
+# CLAP (larger_clap_general) for Core ML — Gridshift
+
+Core ML conversion of [`laion/larger_clap_general`](https://huggingface.co/laion/larger_clap_general), used by Gridshift's smart sample search.
+
+- `GridshiftCLAP.mlpackage.zip`: audio encoder. Input `audio`: fp32 [1, 480000], mono 48 kHz, 10 s (HF `repeatpad` for shorter files). Output `embedding`: fp32 [1, 512], L2-normalized. Log-mel front-end in the graph (matches `ClapFeatureExtractor`), front-end ops fp32, int8 weights (convolutions excepted).
+- `GridshiftCLAPText.mlpackage.zip`: text encoder. Inputs `input_ids`, `attention_mask`: int32 [1, 64] (RoBERTa tokenizer in `tokenizer.tokenizer.zip`). Output `embedding`: fp32 [1, 512], same space.
+
+Validated against the PyTorch reference (`transformers` ClapModel): audio cosine ≥ 0.9995 per sample, pairwise structure Δ ≤ 0.012; text cosine ≥ 0.999. SHA-256s are in `manifest.json` and `text_manifest.json`.
+
+Upstream: LAION-CLAP, Apache-2.0. Wu et al., 2022, *Large-scale Contrastive Language-Audio Pretraining with Feature Fusion and Keyword-to-Caption Augmentation* (arXiv:2211.06687).

manifest.json

@@ -0,0 +1,46 @@
+{
+  "name": "clap_general_coreml_v2",
+  "version": "2.0.0",
+  "format": "coreml",
+  "backend": "coreml",
+  "sample_rate_hz": 48000,
+  "audio_samples": 480000,
+  "embed_dim": 512,
+  "inputs": [
+    {
+      "name": "audio",
+      "layout": "BT",
+      "dtype": "f32",
+      "shape": [
+        1,
+        480000
+      ]
+    }
+  ],
+  "outputs": [
+    {
+      "name": "embedding",
+      "layout": "BD",
+      "dtype": "f32",
+      "shape": [
+        1,
+        512
+      ],
+      "normalization": "l2"
+    }
+  ],
+  "artifacts": [
+    {
+      "file": "GridshiftCLAP.mlpackage.zip",
+      "sha256": "5512b6fa4aab44b2a08dcb505ae31521b20f849025b2c8a941e81fc4e543f21e",
+      "size_bytes": 66006191,
+      "url": "https://huggingface.co/gridshiftstudio/clap-general-coreml/resolve/main/GridshiftCLAP.mlpackage.zip"
+    }
+  ],
+  "entry": "GridshiftCLAP.mlpackage.zip",
+  "upstream": {
+    "source": "laion/larger_clap_general",
+    "license": "Apache-2.0",
+    "paper": "https://arxiv.org/abs/2211.06687"
+  }
+}

text_manifest.json

@@ -0,0 +1,71 @@
+{
+  "name": "clap_general_text_coreml_v2",
+  "version": "2.0.0",
+  "format": "coreml",
+  "backend": "coreml",
+  "max_seq_len": 64,
+  "embed_dim": 512,
+  "tokenizer_meta": {
+    "pad_token_id": 1,
+    "cls_token_id": 0,
+    "sep_token_id": 2,
+    "mask_token_id": 50264,
+    "unk_token_id": 3,
+    "max_seq_len": 64,
+    "model_max_length": 512
+  },
+  "inputs": [
+    {
+      "name": "input_ids",
+      "layout": "BS",
+      "dtype": "i32",
+      "shape": [
+        1,
+        64
+      ]
+    },
+    {
+      "name": "attention_mask",
+      "layout": "BS",
+      "dtype": "i32",
+      "shape": [
+        1,
+        64
+      ]
+    }
+  ],
+  "outputs": [
+    {
+      "name": "embedding",
+      "layout": "BD",
+      "dtype": "f32",
+      "shape": [
+        1,
+        512
+      ],
+      "normalization": "l2",
+      "space": "clap_shared"
+    }
+  ],
+  "artifacts": [
+    {
+      "file": "GridshiftCLAPText.mlpackage.zip",
+      "sha256": "7f2d1bababce0e0fc74f6934ea46b5e5dcfd99167213600276f60285ec87257d",
+      "size_bytes": 112265825,
+      "url": "https://huggingface.co/gridshiftstudio/clap-general-coreml/resolve/main/GridshiftCLAPText.mlpackage.zip"
+    },
+    {
+      "file": "tokenizer.tokenizer.zip",
+      "sha256": "d0631f462b3e90d074824177c32572b9d4a43a8bf503ec9e80b873f4ae9d0e61",
+      "size_bytes": 1187601,
+      "url": "https://huggingface.co/gridshiftstudio/clap-general-coreml/resolve/main/tokenizer.tokenizer.zip"
+    }
+  ],
+  "entry_model": "GridshiftCLAPText.mlpackage.zip",
+  "entry_tokenizer": "tokenizer.tokenizer.zip",
+  "upstream": {
+    "source": "laion/larger_clap_general",
+    "license": "Apache-2.0",
+    "paper": "https://arxiv.org/abs/2211.06687"
+  }
+}

tokenizer.tokenizer.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0631f462b3e90d074824177c32572b9d4a43a8bf503ec9e80b873f4ae9d0e61
+size 1187601