olive-recipes/microsoft-table-transformer-detection/QNN/ttd_config.json at main · microsoft/olive-recipes · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
{
    "input_model": {
        "type": "HfModel",
        "model_path": "microsoft/table-transformer-detection",
        "task": "object-detection",
        "io_config": {
            "input_names": [ "input" ],
            "input_shapes": [ [ 1, 3, 800, 800 ] ],
            "output_names": [ "logits", "pred_boxes" ]
        }
    },
    "systems": {
        "qnn_system": {
            "type": "LocalSystem",
            "accelerators": [ { "device": "npu", "execution_providers": [ "QNNExecutionProvider" ] } ]
        }
    },
    "data_configs": [
        {
            "name": "tablebank_data_config",
            "user_script": "ttd.py",
            "load_dataset_config": { "type": "dataset_load", "data_dir": "tablebank_test" },
            "post_process_data_config": { "type": "dataset_post_process" }
        }
    ],
    "evaluators": {
        "common_evaluator": {
            "metrics": [
                {
                    "name": "accuracy",
                    "type": "custom",
                    "data_config": "tablebank_data_config",
                    "sub_types": [
                        { "name": "mean_of_max_iou", "priority": 1, "higher_is_better": true },
                        { "name": "f1_score", "priority": 2, "higher_is_better": true }
                    ],
                    "user_config": { "user_script": "ttd.py", "metric_func": "evaluate" }
                },
                {
                    "name": "latency",
                    "type": "latency",
                    "data_config": "tablebank_data_config",
                    "sub_types": [ { "name": "avg", "priority": 3 } ]
                }
            ]
        }
    },
    "passes": {
        "conversion": { "device": "cpu", "type": "OnnxConversion", "target_opset": 17 },
        "QNNPreprocess": { "type": "QNNPreprocess", "fuse_layernorm": true },
        "OnnxQuantization": {
            "type": "OnnxQuantization",
            "data_config": "tablebank_data_config",
            "activation_type": "uint16",
            "precision": "uint8",
            "calibrate_method": "MinMax",
            "quant_preprocess": true
        }
    },
    "evaluator": "common_evaluator",
    "evaluate_input_model": false,
    "host": "qnn_system",
    "target": "qnn_system",
    "output_dir": "results/ttd"
}