omniparser-v2-coreml / runtime_config.json
1of2's picture
OmniParser v2 Core ML bundle (icon_detect_v3 + Florence-2 captioner)
7ffdb35 verified
Raw History Blame Contribute Delete
7.82 kB
{
"app_runtime_metrics": {},
"bundle_id": "local_unverified_IconDetectV3",
"caption_boundary_tradeoff": {
"available": false
},
"caption_runtime": {
"crop_size_before_processor": [
64,
64
],
"generation": {
"decode": {
"skip_special_tokens": true,
"trim_whitespace": true
},
"decoder_prefix_length": 20,
"decoder_start_token_id": 2,
"eos_token_id": 2,
"forced_bos_token_id": 0,
"forced_eos_token_id": 2,
"max_new_tokens": 20,
"no_repeat_ngram_size": 3,
"pad_prefix": true,
"pad_token_id": 1,
"selection": "greedy_argmax_after_generation_constraints"
},
"image_preprocessing": {
"do_convert_rgb": null,
"do_normalize": true,
"do_rescale": true,
"do_resize": true,
"image_mean": [
0.485,
0.456,
0.406
],
"image_std": [
0.229,
0.224,
0.225
],
"rescale_factor": 0.00392156862745098,
"size": {
"height": 768,
"width": 768
}
},
"preprocessor_config": "models/caption/tokenizer/preprocessor_config.json",
"processor_source_model_id": "microsoft/Florence-2-base",
"prompt": {
"attention_mask": [
1,
1,
1,
1,
1,
1,
1,
1
],
"input_ids": [
0,
2264,
473,
5,
2274,
6190,
116,
2
],
"pixel_values_dtype": "float32",
"pixel_values_shape": [
1,
3,
768,
768
],
"prompt": "<CAPTION>"
},
"tokenizer_config": "models/caption/tokenizer/tokenizer_config.json",
"tokenizer_json": "models/caption/tokenizer/tokenizer.json",
"vocab_json": "models/caption/tokenizer/vocab.json"
},
"compute_units": "cpu-and-ne",
"detector_postprocessing": {
"confidence_threshold": 0.05,
"max_detections": 300,
"nms_iou_threshold": 0.1,
"remove_overlap_iou_threshold": 0.7
},
"energy": {},
"facade_runtime_metrics": {},
"hard_ane_blocker": {
"ane_dominant_bundle": true,
"available": true,
"evidence_scope": "current_copied_package_hashes",
"expected_blocker": {
"allowed_non_neural_engine_preferred_count": 0,
"component": "caption_encoder"
},
"hard_ane_shippable": true,
"observed_blocker": {
"non_neural_engine_preferred_count": 0,
"operation": null
},
"passed": true,
"source": "manifest.placement_evidence",
"source_sha256": "8b29c6f27756f1d91d426ab286b9dc88c9ea71919840c90c7eb9895dc2ae2691",
"state": "hard_ane_resolved"
},
"minimum_macos": "15.0",
"models": {
"caption_decoder": {
"inputs": {
"decoder_inputs_embeds": {
"dtype": "float16",
"shape": [
1,
20,
768
]
},
"encoder_hidden_states": {
"dtype": "float16",
"shape": [
1,
585,
768
]
}
},
"output": {
"name": "logits",
"shape": [
1,
20,
51289
]
},
"path": "models/caption/FlorenceDecoderPrefixEmbeds.mlpackage",
"placement": {
"first_non_neural_engine_preferred": null,
"hard_ane": true,
"non_neural_engine_preferred_count": 0,
"planned_operation_count": 315
}
},
"caption_encoder": {
"inputs": {
"pixel_values": {
"dtype": "float16",
"layout": "NCHW",
"shape": [
1,
3,
768,
768
]
},
"prompt_embeds": {
"dtype": "float16",
"shape": [
1,
8,
768
]
}
},
"output": {
"dtype": "float16",
"name": "encoder_hidden_states",
"shape": [
1,
585,
768
]
},
"path": "models/caption/FlorenceEncoderContextEmbeds.mlpackage",
"placement": {
"first_non_neural_engine_preferred": null,
"hard_ane": true,
"non_neural_engine_preferred_count": 0,
"planned_operation_count": 1042
}
},
"caption_shared_embedding": {
"dtype": "float16",
"path": "models/caption/shared_embedding_fp16.npy",
"runtime_policy": "memory_map_read_only",
"shape": [
51289,
768
]
},
"detector": {
"default_function": "detect_640",
"family": "icon_detect_v3",
"functions": [
{
"image_size": 640,
"input": {
"dtype": "float16",
"layout": "NCHW",
"name": "image",
"shape": [
1,
3,
640,
640
]
},
"name": "detect_640",
"output": {
"name": "predictions",
"shape": [
1,
5,
8400
]
}
},
{
"image_size": 1280,
"input": {
"dtype": "float16",
"layout": "NCHW",
"name": "image",
"shape": [
1,
3,
1280,
1280
]
},
"name": "detect_1280",
"output": {
"name": "predictions",
"shape": [
1,
5,
33600
]
}
}
],
"license": "MIT",
"output_layout": "[1, 5, anchors]: class logit, then l, t, r, b in grid units",
"path": "models/detector/IconDetectV3.mlpackage",
"placement": {
"first_non_neural_engine_preferred": null,
"hard_ane": true,
"non_neural_engine_preferred_count": 0,
"planned_operation_count": 1594
},
"preprocessing": {
"color_space": "RGB",
"letterbox": "square_center_floor",
"pad_value": 114,
"resample": "pillow_lanczos",
"value_range": "0..1"
},
"source_revision": "f55d0750e5b94db2125ef0b45b0fa4a85ddc59b4",
"source_sha256": "11c6cbb77f22569fab22d86c76407a83ec81ab89dbfe28279854822d6e3fb00c",
"strides": [
8,
16,
32
]
}
},
"policy": {
"ane_as_much_as_possible_without_behavior_change": "unverified",
"ane_dominant_faithful_candidate": false,
"blocker": {
"component": null,
"interpretation": "fresh placement evidence reports zero non-ANE executable ops",
"operation": null
},
"caption_encoder_allowed_cpu_fallbacks": 0,
"hard_ane_shippable": true,
"literal_hard_ane": "satisfied"
},
"runtime_cache": {
"available": false,
"failure_reason": "missing report",
"source": "artifacts/release/coreml_runtime_cache_report.json"
},
"runtime_metrics": {},
"schema_version": 3,
"storage_footprint": {
"available": true,
"caption_decoder_bytes": 192475683,
"caption_encoder_bytes": 269990610,
"caption_model_bytes": 541246325,
"content_bytes_excluding_runtime_config": 660781322,
"detector_bytes": 117076927,
"model_bytes": 658323252,
"policy": {
"compiled_mlmodelc_in_bundle_allowed": false,
"max_caption_decoder_bytes": 251658240,
"max_caption_encoder_bytes": 335544320,
"max_detector_bytes": 83886080,
"max_model_bytes": 681574400,
"max_shared_embedding_bytes": 100663296,
"max_tokenizer_bytes": 4194304,
"ship_mlpackage_only": true
},
"runtime_config_self_size_excluded": true,
"shared_embedding_bytes": 78780032,
"tokenizer_bytes": 2456587
},
"target": "macos"
}