Upload 58 files

cond_step.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32a2f7ca50a9b4d14e44b26fdad9187cff107c385404dfd7c4ddaf78e9c8f59d
+size 133877

cond_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b94b9ea4bc9a1e67863f7f388778482d9f2fa58c852f2b3419d7c6b803acd0f5
+size 265188352

cond_step.mlpackage/Manifest.json

@@ -0,0 +1,18 @@
+{
+    "fileFormatVersion": "1.0.0",
+    "itemInfoEntries": {
+        "28904155-48D7-4BF7-8ABE-BCD892E54E10": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Specification",
+            "name": "model.mlmodel",
+            "path": "com.apple.CoreML/model.mlmodel"
+        },
+        "BE1770FC-0E0B-4C68-B3A6-5F71B473B424": {
+            "author": "com.apple.CoreML",
+            "description": "CoreML Model Weights",
+            "name": "weights",
+            "path": "com.apple.CoreML/weights"
+        }
+    },
+    "rootModelIdentifier": "28904155-48D7-4BF7-8ABE-BCD892E54E10"
+}

constants_bin/alba_audio_prompt.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b2429b1cce77de299cfe79900a7662b342e58674e659a17892fb08103e389fb
+size 512000

constants_bin/alba_voice_cache/cache0.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99392e82f7a10cfdae1e82d5efe4119ab13742d0aa6d240c50d3b11658977e57
+size 1638400

constants_bin/alba_voice_cache/cache1.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a1e6e03d989baf84d9659ea5eaff60e07cf32850d8a3111bda2d99feb5c99f9
+size 1638400

constants_bin/alba_voice_cache/cache2.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b62d9ec911282a3441e936ef08e480008762c4394d9429f194ba4f8178b94125
+size 1638400

constants_bin/alba_voice_cache/cache3.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10de79d9ff8e4dbbb561155d415bab9977fbab94979d086cc2b134ae8f40cc5a
+size 1638400

constants_bin/alba_voice_cache/cache4.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cd23f11e060fce8cb05449b362f9a114c9eac9034f33a616169bf779f3eafaa
+size 1638400

constants_bin/alba_voice_cache/cache5.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2a735340c2d08b663ec10e61dd8356e6f24cd7493ad8356d12eb8b763829508
+size 1638400

constants_bin/alba_voice_cache/position0.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/alba_voice_cache/position1.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/alba_voice_cache/position2.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/alba_voice_cache/position3.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/alba_voice_cache/position4.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/alba_voice_cache/position5.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5bec6bab2611a6b777c2956be5365f49fa3e2769832d8046d96e93d80cb8f1
+size 4

constants_bin/bos_emb.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e66fe9dfbbb7d543ed6271144191112364481e1f8e1507e6ab51f412b5348bd4
+size 128

constants_bin/emb_mean.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa999d96639a54952c213d9362b418d724eb2d0b23274733b1b1b78ffcb29468
+size 128

constants_bin/emb_std.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:196d4ad8655bde0b2f9d3ac1ae83af34f27b46f6d71c32bcdc2e299fa58595ee
+size 128

constants_bin/manifest.json

@@ -0,0 +1,379 @@
+{
+  "bos_emb": {
+    "shape": [
+      32
+    ],
+    "dtype": "float32",
+    "bytes": 128
+  },
+  "emb_mean": {
+    "shape": [
+      32
+    ],
+    "dtype": "float32",
+    "bytes": 128
+  },
+  "emb_std": {
+    "shape": [
+      32
+    ],
+    "dtype": "float32",
+    "bytes": 128
+  },
+  "quantizer_weight": {
+    "shape": [
+      512,
+      32,
+      1
+    ],
+    "dtype": "float32",
+    "bytes": 65536
+  },
+  "text_embed_table": {
+    "shape": [
+      4001,
+      1024
+    ],
+    "dtype": "float32",
+    "bytes": 16388096
+  },
+  "mimi_init_state": {
+    "upsample_partial": {
+      "shape": [
+        1,
+        512,
+        16
+      ],
+      "dtype": "float32",
+      "bytes": 32768
+    },
+    "attn0_cache": {
+      "shape": [
+        2,
+        1,
+        8,
+        256,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1048576
+    },
+    "attn0_offset": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "attn0_end_offset": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "attn1_cache": {
+      "shape": [
+        2,
+        1,
+        8,
+        256,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1048576
+    },
+    "attn1_offset": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "attn1_end_offset": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "conv0_prev": {
+      "shape": [
+        1,
+        512,
+        6
+      ],
+      "dtype": "float32",
+      "bytes": 12288
+    },
+    "conv0_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "convtr0_partial": {
+      "shape": [
+        1,
+        256,
+        6
+      ],
+      "dtype": "float32",
+      "bytes": 6144
+    },
+    "res0_conv0_prev": {
+      "shape": [
+        1,
+        256,
+        2
+      ],
+      "dtype": "float32",
+      "bytes": 2048
+    },
+    "res0_conv0_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "res0_conv1_prev": {
+      "shape": [
+        1,
+        128,
+        0
+      ],
+      "dtype": "float32",
+      "bytes": 0
+    },
+    "res0_conv1_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "convtr1_partial": {
+      "shape": [
+        1,
+        128,
+        5
+      ],
+      "dtype": "float32",
+      "bytes": 2560
+    },
+    "res1_conv0_prev": {
+      "shape": [
+        1,
+        128,
+        2
+      ],
+      "dtype": "float32",
+      "bytes": 1024
+    },
+    "res1_conv0_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "res1_conv1_prev": {
+      "shape": [
+        1,
+        64,
+        0
+      ],
+      "dtype": "float32",
+      "bytes": 0
+    },
+    "res1_conv1_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "convtr2_partial": {
+      "shape": [
+        1,
+        64,
+        4
+      ],
+      "dtype": "float32",
+      "bytes": 1024
+    },
+    "res2_conv0_prev": {
+      "shape": [
+        1,
+        64,
+        2
+      ],
+      "dtype": "float32",
+      "bytes": 512
+    },
+    "res2_conv0_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "res2_conv1_prev": {
+      "shape": [
+        1,
+        32,
+        0
+      ],
+      "dtype": "float32",
+      "bytes": 0
+    },
+    "res2_conv1_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "conv_final_prev": {
+      "shape": [
+        1,
+        64,
+        2
+      ],
+      "dtype": "float32",
+      "bytes": 512
+    },
+    "conv_final_first": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    }
+  },
+  "alba_voice_cache": {
+    "cache0": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position0": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "cache1": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position1": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "cache2": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position2": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "cache3": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position3": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "cache4": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position4": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    },
+    "cache5": {
+      "shape": [
+        2,
+        1,
+        200,
+        16,
+        64
+      ],
+      "dtype": "float32",
+      "bytes": 1638400
+    },
+    "position5": {
+      "shape": [
+        1
+      ],
+      "dtype": "float32",
+      "bytes": 4
+    }
+  },
+  "alba_audio_prompt": {
+    "shape": [
+      1,
+      125,
+      1024
+    ],
+    "dtype": "float32",
+    "bytes": 512000
+  },
+  "tokenizer": {
+    "format": "sentencepiece_protobuf",
+    "bytes": 59339
+  }
+}

constants_bin/mimi_init_state/attn0_cache.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30e14955ebf1352266dc2ff8067e68104607e750abb9d3b36582b8af909fcb58
+size 1048576

constants_bin/mimi_init_state/attn0_end_offset.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df3f619804a92fdb4057192dc43dd748ea778adc52bc498ce80524c014b81119
+size 4

constants_bin/mimi_init_state/attn0_offset.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df3f619804a92fdb4057192dc43dd748ea778adc52bc498ce80524c014b81119
+size 4

constants_bin/mimi_init_state/attn1_cache.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30e14955ebf1352266dc2ff8067e68104607e750abb9d3b36582b8af909fcb58
+size 1048576

constants_bin/mimi_init_state/attn1_end_offset.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df3f619804a92fdb4057192dc43dd748ea778adc52bc498ce80524c014b81119
+size 4

constants_bin/mimi_init_state/attn1_offset.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df3f619804a92fdb4057192dc43dd748ea778adc52bc498ce80524c014b81119
+size 4

constants_bin/mimi_init_state/conv0_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/conv0_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3cc103136423a57975750907ebc1d367e2985ac6338976d4d5a439f50323f4a
+size 12288

constants_bin/mimi_init_state/conv_final_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/conv_final_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:076a27c79e5ace2a3d47f9dd2e83e4ff6ea8872b3c2218f66c92b89b55f36560
+size 512

constants_bin/mimi_init_state/convtr0_partial.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd9243e1ba57263ed469c3bdbd7ade6ec5254e7ed924a9f5737fa44749933cc0
+size 6144

constants_bin/mimi_init_state/convtr1_partial.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ce8ba8e726ee8925e6560d86ac35be1097691d1cfac888e6bd20e804ea9eb15
+size 2560

constants_bin/mimi_init_state/convtr2_partial.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef
+size 1024

constants_bin/mimi_init_state/res0_conv0_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res0_conv0_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5a00aa9991ac8a5ee3109844d84a55583bd20572ad3ffcd42792f3c36b183ad
+size 2048

constants_bin/mimi_init_state/res0_conv1_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res0_conv1_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
+size 0

constants_bin/mimi_init_state/res1_conv0_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res1_conv0_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef
+size 1024

constants_bin/mimi_init_state/res1_conv1_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res1_conv1_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
+size 0

constants_bin/mimi_init_state/res2_conv0_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res2_conv0_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:076a27c79e5ace2a3d47f9dd2e83e4ff6ea8872b3c2218f66c92b89b55f36560
+size 512

constants_bin/mimi_init_state/res2_conv1_first.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e00e5eb9444182f352323374ef4e08ebcb784725fdd4fd612d7730540b3e0c8c
+size 4

constants_bin/mimi_init_state/res2_conv1_prev.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
+size 0

constants_bin/mimi_init_state/upsample_partial.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c35020473aed1b4642cd726cad727b63fff2824ad68cedd7ffb73c7cbd890479
+size 32768

constants_bin/quantizer_weight.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1efa734d33fde572f43cc050756960698ca3310723364f5bb6959ddff829d2d
+size 65536

constants_bin/text_embed_table.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08100a5e7ac7a90ade31d38d4ec379309c136b2d9c9fa6709b3d3ffc61d751ce
+size 16388096

constants_bin/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d461765ae179566678c93091c5fa6f2984c31bbe990bf1aa62d92c64d91bc3f6
+size 59339

flow_decoder_v2.mlpackage/Data/com.apple.CoreML/model.mlmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:298d7812b93dd74f3e97ed622c929627895ae7f23ef41b6077eb984580a84eca
+size 43299