Upload 3 files

README.md

@@ -1,3 +1,35 @@
 ---
 license: cc-by-nc-4.0
+library_name: mlx-audio
+tags:
+- mlx
+- audio
+- audio-source-separation
+base_model: sarulab-speech/DialogueSidon
 ---
+
+# DialogueSidon
+
+Two-speaker dialogue separation and restoration, with 24 kHz mono stems.
+Converted from [sarulab-speech/DialogueSidon](https://huggingface.co/sarulab-speech/DialogueSidon/tree/d43d7478402a5527136c6733c3f4359c37b312ab) for use with [mlx-audio](https://github.com/Blaizzy/mlx-audio).
+
+```python
+from mlx_audio.sts import load
+from mlx_audio.audio_io import write
+
+model = load("mlx-community/DialogueSidon")
+result = model.separate("dialogue.wav", num_steps=30, seed=0)
+for i, speaker in enumerate(result.speakers, 1):
+    write(f"speaker_{i}.wav", speaker, result.sample_rate)
+```
+
+The two channels are anonymous speaker slots. Long recordings use overlapping
+chunks and waveform correlation to maintain channel assignment. Consistency
+across silence is not guaranteed. This is an offline generative model, so the
+stems need not sum to the mixture. The default chunk and overlap are 20 and 5
+seconds; use `chunk_seconds=None` for whole-file inference when memory permits.
+
+The checkpoint retains **CC-BY-NC-4.0**. Original model by Wataru Nakata,
+Yuki Saito, Kazuki Yamauchi, Emiru Tsunoo, and Hiroshi Saruwatari (SaruLab).
+See the [original model card](https://huggingface.co/sarulab-speech/DialogueSidon) and
+[Sidon implementation](https://github.com/sarulab-speech/Sidon).

config.json

@@ -0,0 +1,170 @@
+{
+  "model_type": "dialogue_sidon",
+  "sample_rate": 24000,
+  "latent_dim": 32,
+  "encoder": {
+    "hidden_size": 1024,
+    "intermediate_size": 4096,
+    "num_hidden_layers": 13,
+    "num_attention_heads": 16,
+    "feature_projection_input_dim": 160,
+    "conv_depthwise_kernel_size": 31,
+    "left_max_position_embeddings": 64,
+    "right_max_position_embeddings": 8,
+    "layer_norm_eps": 1e-05
+  },
+  "diffusion": {
+    "hidden_size": 768,
+    "num_layers": 8,
+    "num_heads": 12,
+    "ffn_ratio": 4.0,
+    "frequency_embedding_size": 256,
+    "num_train_timesteps": 1000,
+    "beta_start": 0.0001,
+    "beta_end": 0.02,
+    "prediction_type": "v_prediction"
+  },
+  "decoder_channels": 1536,
+  "decoder_rates": [
+    8,
+    5,
+    4,
+    3
+  ],
+  "latent_norm_initialized": true,
+  "latent_norm_mean": [
+    0.1416097730398178,
+    0.07175948470830917,
+    0.16445186734199524,
+    -0.0158994160592556,
+    -0.03792296350002289,
+    0.1739553064107895,
+    0.20609763264656067,
+    -0.0773492380976677,
+    -0.0958249643445015,
+    -0.14557723701000214,
+    -0.38258203864097595,
+    0.08845309168100357,
+    -0.029827825725078583,
+    0.04720787703990936,
+    -0.20060305297374725,
+    0.11549390852451324,
+    0.06303639709949493,
+    0.024705005809664726,
+    -0.0034871164243668318,
+    -0.09642789512872696,
+    0.21524178981781006,
+    0.13936381042003632,
+    0.24526667594909668,
+    0.2592932879924774,
+    -0.3074306547641754,
+    0.14334627985954285,
+    0.07093678414821625,
+    0.5708937048912048,
+    0.006343616638332605,
+    0.09566544741392136,
+    0.5273311734199524,
+    0.1451909840106964,
+    -0.006651828996837139,
+    0.10430128127336502,
+    0.11032214015722275,
+    0.012842725962400436,
+    -0.024605553597211838,
+    0.12466253340244293,
+    0.12605680525302887,
+    -0.02444394864141941,
+    -0.1059999093413353,
+    -0.1217033863067627,
+    -0.3549865186214447,
+    0.09900717437267303,
+    0.01073773019015789,
+    0.10321037471294403,
+    -0.11944245547056198,
+    0.09993556141853333,
+    0.01015459280461073,
+    0.03638560324907303,
+    -0.0326518714427948,
+    -0.13041582703590393,
+    0.12432265281677246,
+    0.09911591559648514,
+    0.16202199459075928,
+    0.20371796190738678,
+    -0.2879335582256317,
+    0.067603699862957,
+    0.010356826707720757,
+    0.45639586448669434,
+    0.05465073511004448,
+    0.014600546099245548,
+    0.4818445146083832,
+    0.09414693713188171
+  ],
+  "latent_norm_std": [
+    0.4863946735858917,
+    0.39176884293556213,
+    0.5244080424308777,
+    0.31121498346328735,
+    0.38742581009864807,
+    0.39328351616859436,
+    0.353639155626297,
+    0.43606036901474,
+    0.40308451652526855,
+    0.45672574639320374,
+    0.43666139245033264,
+    0.31756022572517395,
+    0.44575342535972595,
+    0.3706947863101959,
+    0.5126213431358337,
+    0.3701273798942566,
+    0.4332931637763977,
+    0.34073886275291443,
+    0.36883777379989624,
+    0.3678142726421356,
+    0.421845406293869,
+    0.4466390907764435,
+    0.4708837568759918,
+    0.41468745470046997,
+    0.5738605856895447,
+    0.4226427376270294,
+    0.3899722695350647,
+    0.7289687395095825,
+    0.4073668420314789,
+    0.4150795340538025,
+    0.5346329212188721,
+    0.3559742569923401,
+    0.5018235445022583,
+    0.40256914496421814,
+    0.5634673833847046,
+    0.3569527864456177,
+    0.39661917090415955,
+    0.4181853234767914,
+    0.40074092149734497,
+    0.4378126561641693,
+    0.45497414469718933,
+    0.5129145383834839,
+    0.4792986810207367,
+    0.3506719768047333,
+    0.4913414418697357,
+    0.4084736704826355,
+    0.5545224547386169,
+    0.3966507315635681,
+    0.47008392214775085,
+    0.3715076744556427,
+    0.40553003549575806,
+    0.368908166885376,
+    0.4457831382751465,
+    0.4751212000846863,
+    0.4867909550666809,
+    0.4290574789047241,
+    0.5897377729415894,
+    0.4735560417175293,
+    0.44096100330352783,
+    0.790938675403595,
+    0.4382975697517395,
+    0.4624195098876953,
+    0.5593090057373047,
+    0.3772125244140625
+  ],
+  "source_repo": "sarulab-speech/DialogueSidon",
+  "source_revision": "d43d7478402a5527136c6733c3f4359c37b312ab",
+  "torch_dtype": "float32"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:196bf68cd7539da55f7f798467e566bd2c11415c3f3a9925e322c265eeac5e96
+size 1776912148