coreml-mop-insta-v1-dmd-6bit / PROVENANCE.json
Devserhii's picture
Record full Core ML LCM validation on Mac GPU and Neural Engine
262047d verified
Raw History Blame Contribute Delete
5.42 kB
{
"schemaVersion": 1,
"modelId": "mop_insta_v1_dmd_coreml",
"family": "SDXL 1.0",
"variant": "DMD",
"resolution": [
1024,
1024
],
"source": {
"author": "GBRX",
"modelName": "MoP (Mixture of Perverts) DMD",
"versionName": "Insta v1.0 DMD",
"modelId": 1854124,
"modelVersionId": 2811401,
"fileId": 2697274,
"filename": "mopMixtureOfPerverts_instaV10DMD.safetensors",
"sizeBytes": 6938083554,
"sha256": "fd05f05d92afc82bf3b1c3992c10e116a73b7d26ba464cc3510f530b16aa6aac",
"url": "https://civitai.com/models/1854124?modelVersionId=2811401",
"licenseURL": "https://civitai.com/models/license/2811401"
},
"configuration": {
"repository": "stabilityai/stable-diffusion-xl-base-1.0",
"revision": "462165984030d82259a11f4367a4eed129e94a7b",
"configManifestSha256": "dc20c30615ebc3100a8982beb1a71d7ba4a8d9c0aa22435f0aaea74ca1fb0415",
"note": "Configuration and tokenizer only; trainable UNet, VAE, and used text-encoder weights are from the authenticated MoP checkpoint."
},
"textEncoderRepair": {
"reason": "Exactly six tensors in the unused final CLIP-L transformer layer contain non-finite values. SDXL uses the penultimate CLIP-L hidden state for conditioning.",
"affectedLayer": "text_model.encoder.layers.11",
"affectedTensorCount": 6,
"replacedNonFiniteValueCount": 7077888,
"replacementRepository": "stabilityai/stable-diffusion-xl-base-1.0",
"replacementRevision": "462165984030d82259a11f4367a4eed129e94a7b",
"replacementFile": "text_encoder/model.fp16.safetensors",
"replacementSizeBytes": 246144152,
"replacementSha256": "660c6f5b1abae9dc498ac2d21e1347d2abdb0cf6c0c0c8576cd796491d9a6cdd",
"scope": "Only non-finite values in the unused final CLIP-L layer were replaced. All used conditioning layers remain from MoP."
},
"conversion": {
"appleMLStableDiffusionRevision": "e12202c1f6405b83918b58a5d097cd61e3e1f702",
"localUnetConverterSha256": "a5a174f86a52309a580924584c72ef79259664d37fc01ec8bcd2037f850e2e40",
"localTextEncoderConverterSha256": "5bcf45966a7cdf130f91ddf85d6d583228223735d899839754b2d0b3f71b8947",
"coremltools": "8.0",
"diffusers": "0.30.2",
"torch": "2.4.0",
"transformers": "4.44.2",
"safetensors": "0.4.5",
"minimumDeploymentTarget": "iOS 17",
"attentionImplementation": "SPLIT_EINSUM",
"unet": {
"quantizationBits": 6,
"method": "k-means palettization",
"split": true,
"batchSize": 2,
"inputShape": [
2,
4,
128,
128
],
"outputShape": [
2,
4,
128,
128
]
},
"textEncoder": {
"precision": "FP16",
"source": "MoP except documented unused-layer repair"
},
"textEncoder2": {
"quantizationBits": 8,
"source": "MoP"
},
"vae": {
"source": "MoP",
"decoderIncluded": true,
"encoderIncludedForFaceDetail": true
}
},
"componentWeightBytes": {
"TextEncoder.mlmodelc": 246145536,
"TextEncoder2.mlmodelc": 758591232,
"UnetChunk1.mlmodelc": 968067264,
"UnetChunk2.mlmodelc": 959821632,
"VAEDecoder.mlmodelc": 197977216,
"VAEEncoder.mlmodelc": 136668992
},
"generation": {
"scheduler": "LCM",
"defaultSteps": 10,
"recommendedSteps": [
8,
12
],
"defaultCFG": 1.0,
"recommendedCFG": [
1.0,
1.3
],
"width": 1024,
"height": 1024
},
"license": {
"name": "CreativeML Open RAIL++-M with MoP Attachment B",
"url": "https://civitai.com/models/license/2811401",
"file": "LICENSE"
},
"validation": {
"sourceDiffusersGeneration": true,
"sourceSettings": {
"scheduler": "LCMScheduler",
"steps": 10,
"cfg": 1.0,
"resolution": [
1024,
1024
],
"seed": 1568
},
"fp16UnetPyTorchToCoreMLPSNRdB": 45.5,
"textEncoderPyTorchToCoreMLPSNRdB": 90.8,
"textEncoder2PyTorchToCoreMLPSNRdB": 69.3,
"vaeDecoderPyTorchToCoreMLPSNRdB": 108.5,
"vaeEncoderPyTorchToCoreMLPSNRdB": 101.2,
"compiledComponentShapesVerified": true,
"allWeightFilesUnder1GiB": true,
"swiftPipelineMacCPUGPU": {
"success": true,
"scheduler": "LCM",
"steps": 10,
"cfg": 1.0,
"seed": 1568,
"resolution": [
1024,
1024
],
"parrotPromptOutputSha256": "df79e890fe05a73f215a93c58236df44a64f34d4ac5e0593120647c61a3d3a79",
"portraitPromptOutputSha256": "a688237e9cc44a7bc07b45cc23aadc0e68909b2b2b2e8e64c6ab8dd41b93f089",
"note": "Both outputs visually inspected: photorealistic, prompt-aligned, no mosaic/noise failure."
},
"sourceDiffusersOutputSha256": "340d6c45d9dfd540e467f68e2ee518dcd86f72b64d77a59e8067cad6ca2ff8d1",
"swiftPipelineMacNeuralEngine": {
"success": true,
"scheduler": "LCM",
"steps": 10,
"cfg": 1.0,
"seed": 1568,
"resolution": [
1024,
1024
],
"parrotPromptOutputSha256": "0517adc2cb243c88e42d115589f817db3110b59449d47465124d780a221941b6",
"portraitPromptOutputSha256": "32bce578f9ba011dd3046e7de20f4e5e2f324948bf4b8b9d3805a6e8c8468d9b",
"note": "Both outputs visually inspected: photorealistic, prompt-aligned, no mosaic/noise failure. The first ANE plan compilation on this Mac took several minutes before inference."
}
}
}