psiddh's picture
Add ExecuTorch WebGPU program for phonon-31m-en
afd4b73 verified
Raw History Blame Contribute Delete
7.27 kB
{
"model": "phonon-31m-en",
"upstream": {
"model": "gradium/phonon-31m-en",
"revision": "52433c46975bfa13144c56204c9f4ee1c5b148aa",
"voices": {
"repo": "gradium/phonon-voices",
"revision": "0ee17205f254b5abc43b266c8a0f215565c352f7"
},
"official": true,
"license": "apache-2.0 (model card front matter and License section; phonon-voices card front matter)",
"files_sha256": {
"gradium/phonon-31m-en:model.safetensors": "8a04cf833cfe69b67c2ec968b2d31244655b9accc84ade9c0ebcc176f2074607",
"gradium/phonon-31m-en:config.json": "7062ea376e6b158543c8abec8e5af67df171c49ad70eb7e2a0fb25c16f265e06",
"gradium/phonon-31m-en:tokenizer.json": "691a00068fa7b989f2f3e5ae5daa2d1a5bb319f2d5360deafcfb19a8bb6e7936",
"gradium/phonon-31m-en:tokenizer.model": "82e4572006d64d483317074c22c368e87779640b6f3a9dcd07b7caae32b030ce",
"gradium/phonon-31m-en:default-voice.safetensors": "f739e8c4c76ee7f92e7c94a82002748c8cd22e99582ac49f7b9b94cd63786e18",
"gradium/phonon-31m-en:README.md": "00f09ecdcd23ed9904f3a503baecc1f0c77515637ce6172b572ed3f50ce50c29",
"gradium/phonon-voices:voices/LMDIYE0Oa377Qh6Z.safetensors": "45c582673cb20edb04ddcca2b5a8865ddfb0a82599efadb269021e87d20fbdd4",
"gradium/phonon-voices:voices.jsonl": "874f34998ce581b94f2b8912aa015639f52b5fd62eba3ba11e1b64dcaed7d802",
"gradium/phonon-voices:README.md": "0186c7d586a9594db1e0283cb68e84e88bbd9bc3c29047cf0083c7483556f3ee"
},
"loader": "recipes/_phonon.py load_weights: model.safetensors (bf16 flow LM, f32 Mimi decoder) -> float32, load_state_dict(strict=True) into a module whose keys equal the checkpoint's"
},
"versions": {
"python": "3.12.15+meta",
"torch": "2.14.1+cpu",
"torchao": "0.18.0+cpu",
"executorch": "1.5.1+cpu",
"executorch_git": "3b60683923245cf472b7323426920e15623ba361",
"transformers": "5.19.0",
"huggingface_hub": "2.2.0"
},
"export": {
"api": "torch.export.export(strict=False) -> to_edge_transform_and_lower(partitioner=[WebGPUPartitioner(...)]) -> to_executorch()",
"partitioner": {
"flow_lm": "WebGPUPartitioner(compile_options={'require_dynamic_shapes': True, 'force_fp16': False}), dynamic S in [1, 128]",
"mimi": "WebGPUPartitioner() (default compile options, static shapes)"
},
"dtype": "float32 (bf16 checkpoint weights converted exactly)",
"quantization": "none",
"model_code": "recipes/_phonon.py FlowLMProgram, MimiStepProgram"
},
"files": {
"phonon_flow_lm.pte": {
"bytes": 111847044,
"sha256": "178698a6c4539ba48c93bd74c4165586c1da4fdd393bd144c27fd09d80021ad7"
},
"phonon_mimi.pte": {
"bytes": 41304196,
"sha256": "d7acdf2d93ba4897e63b58d933c7425f49cf855ecbd1a6f51e273c5dd65bb3cc"
},
"phonon_text_embed.bf16.bin": {
"bytes": 4097024,
"sha256": "7c6f3394fc86b55ed181ff0b9d2472cf0371d3feeac25a5d0f2eed0a3a2eaae1"
},
"phonon_bos_embed.f32.bin": {
"bytes": 2048,
"sha256": "a4464c50596adea27c8f0457124ead461202dc2f67daf3c6dbbc289cf289fd32"
},
"voice_default.f32.bin": {
"bytes": 256000,
"sha256": "d8a8ffc97e46c0330ef6a3842130b95cc5ee71759f66bb54bc2ec0b79fc36e57"
},
"voice_en_us_easton.f32.bin": {
"bytes": 256000,
"sha256": "1d6ec67cbd3a45b2bacf84ff7ddfb184fd9e8467c2204f2eac0f4975c354b66b"
},
"tokenizer.json": {
"bytes": 277967,
"sha256": "691a00068fa7b989f2f3e5ae5daa2d1a5bb319f2d5360deafcfb19a8bb6e7936"
},
"phonon.json": {
"bytes": 8599,
"sha256": "e8fa6604769aeb3b63e95c44c912f7ae635a324ca3793ec62fca017346521750"
}
},
"graph": {
"flow_lm": {
"non_delegated_kernels": {},
"delegates": [
{
"id": "VulkanBackend",
"ops": {
"et_vk.prepack.default": 23,
"sym_size.int": 1,
"et_vk.select_as_symint.default": 1,
"aten.view_copy.default": 106,
"aten.linear.default": 60,
"aten.sub.Tensor": 12,
"et_vk.rms_norm.default": 12,
"aten.add.Tensor": 47,
"aten.slice_copy.Tensor": 38,
"et_vk.apply_rotary_emb_hf.default": 6,
"sdpa_with_kv_cache.default": 6,
"aten.gelu.default": 6,
"aten.select_copy.int": 1,
"aten.native_layer_norm.default": 8,
"aten.sigmoid.default": 7,
"aten.mul.Tensor": 20,
"aten.unsqueeze_copy.default": 1
},
"inputs": 3,
"outputs": 3
}
]
},
"mimi": {
"non_delegated_kernels": {},
"delegates": [
{
"id": "VulkanBackend",
"ops": {
"et_vk.prepack.default": 34,
"aten.mul.Tensor": 18,
"aten.add.Tensor": 32,
"aten.view_copy.default": 30,
"aten.linear.default": 9,
"aten.clone.default": 14,
"aten.unsqueeze_copy.default": 9,
"aten.convolution.default": 12,
"aten.squeeze_copy.dims": 12,
"aten.slice_copy.Tensor": 35,
"aten.cat.default": 21,
"aten.cos.default": 2,
"aten.sin.default": 2,
"aten.permute_copy.default": 12,
"aten.native_layer_norm.default": 4,
"aten.neg.default": 4,
"aten.sub.Tensor": 15,
"aten.bmm.default": 4,
"aten._softmax.default": 2,
"aten.gelu.default": 2,
"aten.relu.default": 10,
"aten.clamp.default": 10,
"aten.exp.default": 10
},
"inputs": 16,
"outputs": 16
}
]
}
},
"webgpu_runtime_checks": {
"op_list": "checked against --runtime-ops",
"flow_lm": {
"problems": [],
"ok": true,
"notes": {
"value_kinds": [
"Bool",
"Double",
"FLOAT32",
"INT32",
"Int",
"IntList",
"Null",
"String",
"SymInt",
"ValueList"
],
"aten.linear input ranks": [
"2"
],
"sdpa_with_kv_cache q/k_cache": [
"((1, 128, 8, 64), (1, 512, 8, 64), 'SymInt')"
]
}
},
"mimi": {
"problems": [],
"ok": true,
"notes": {
"value_kinds": [
"Bool",
"Double",
"FLOAT32",
"Int",
"IntList",
"Null",
"String",
"ValueList"
],
"aten.linear input ranks": [
"2"
],
"aten.convolution input/weight": [
"((1, 128, 1, 480), (128, 64, 1, 8))",
"((1, 128, 482), (64, 128, 3))",
"((1, 128, 96), (256, 128, 1))",
"((1, 256, 1, 96), (256, 128, 1, 10))",
"((1, 256, 98), (128, 256, 3))",
"((1, 32, 1920), (64, 32, 1))",
"((1, 512, 1, 1), (512, 1, 1, 32))",
"((1, 512, 1, 16), (512, 256, 1, 12))",
"((1, 512, 22), (512, 512, 7))",
"((1, 64, 1922), (1, 64, 3))",
"((1, 64, 1922), (32, 64, 3))",
"((1, 64, 480), (128, 64, 1))"
]
}
}
},
"command": "python recipes/phonon-31m-en.py --out out/phonon-31m-en"
}