README.md
4.0 KB · 133 lines · markdown Raw
1 ---
2 library_name: keras
3 license: mit
4 tags:
5 - inverse-design
6 - transmon
7 - hamiltonian
8 - squadds
9 - keras
10 - tabular-regression
11 ---
12
13 # TransmonCross Hamiltonian to Geometry
14
15 Inverse model that predicts TransmonCross geometry parameters from target Hamiltonian values.
16
17 ## Live Serving Surface
18
19 - Space repo: https://huggingface.co/spaces/SQuADDS/squadds-ml-inference-api
20 - Space host: https://squadds-squadds-ml-inference-api.hf.space
21
22 ## Inference Contract
23
24 The deployed artifact uses the same request contract as the SQuADDS ML Space:
25
26 ```json
27 {
28 "model_id": "transmon_cross_hamiltonian_inverse",
29 "inputs": {
30 "qubit_frequency_GHz": 4.85,
31 "anharmonicity_MHz": -205.0
32 },
33 "options": {
34 "include_scaled_outputs": false
35 }
36 }
37 ```
38
39 ## Sample Response
40
41 ```json
42 {
43 "model_id": "transmon_cross_hamiltonian_inverse",
44 "display_name": "TransmonCross Hamiltonian to Geometry",
45 "predictions": [
46 {
47 "design_options.connection_pads.readout.claw_length": 0.00011072495544794947,
48 "design_options.connection_pads.readout.ground_spacing": 4.571595582092414e-06,
49 "design_options.cross_length": 0.0002005973074119538
50 }
51 ],
52 "metadata": {
53 "input_order": [
54 "qubit_frequency_GHz",
55 "anharmonicity_MHz"
56 ],
57 "output_order": [
58 "design_options.connection_pads.readout.claw_length",
59 "design_options.connection_pads.readout.ground_spacing",
60 "design_options.cross_length"
61 ],
62 "input_units": {
63 "qubit_frequency_GHz": "GHz",
64 "anharmonicity_MHz": "MHz"
65 },
66 "output_units": {
67 "design_options.connection_pads.readout.claw_length": "m",
68 "design_options.connection_pads.readout.ground_spacing": "m",
69 "design_options.cross_length": "m"
70 },
71 "num_predictions": 1
72 }
73 }
74 ```
75
76 ## Input and Output Fields
77
78 - Input units: `{"anharmonicity_MHz": "MHz", "qubit_frequency_GHz": "GHz"}`
79 - Output units: `{"design_options.connection_pads.readout.claw_length": "m", "design_options.connection_pads.readout.ground_spacing": "m", "design_options.cross_length": "m"}`
80
81 ## Included Files
82
83 - `model/`: trained Keras checkpoint
84 - `scalers/`: per-column input and output scalers when available
85 - `X_names`: ordered input feature names
86 - output-name file (`y_columns.npy` or csv header source)
87 - `inference_manifest.json`: machine-readable contract for agents and clients
88
89 ## SQuADDS Dataset
90
91 This model is derived from the public SQuADDS dataset and related tooling.
92
93 - Dataset page: https://huggingface.co/datasets/SQuADDS/SQuADDS_DB
94 - Dataset file tree: https://huggingface.co/datasets/SQuADDS/SQuADDS_DB/tree/main
95 - SQuADDS datasets org page: https://huggingface.co/SQuADDS/datasets
96 - SQuADDS homepage: https://lfl-lab.github.io/SQuADDS/
97 - SQuADDS repository: https://github.com/LFL-Lab/SQuADDS
98 - SQuADDS paper: https://doi.org/10.22331/q-2024-09-09-1465
99 - Hugging Face dataset DOI: `10.57967/hf/1582`
100
101 For this model family, the most relevant SQuADDS source data is:
102
103 - `qubit-TransmonCross-cap_matrix`
104
105 ## Citation
106
107 If you use SQuADDS data or this ML workflow in research, please cite:
108
109 ```bibtex
110 @article{Shanto2024squaddsvalidated,
111 doi = {10.22331/q-2024-09-09-1465},
112 url = {https://doi.org/10.22331/q-2024-09-09-1465},
113 title = {{SQ}u{ADDS}: {A} validated design database and simulation workflow for superconducting qubit design},
114 author = {Shanto, Sadman and Kuo, Andre and Miyamoto, Clark and Zhang, Haimeng and Maurya, Vivek and Vlachos, Evangelos and Hecht, Malida and Shum, Chung Wa and Levenson-Falk, Eli},
115 journal = {{Quantum}},
116 volume = {8},
117 pages = {1465},
118 month = sep,
119 year = {2024}
120 }
121 ```
122
123 ## Acknowledgments
124
125 We gratefully acknowledge this collaboration for developing the model: Taylor Patti, Nicola Pancotti, Enectali Figueroa-Feliciano, Sara Sussman, Abhishek Chakraborty, Olivia Seidel, Firas Abouzahr, Eli Levenson-Falk and Sadman Ahmed Shanto.
126
127 Special thanks to Olivia Seidel and Firas Abouzahr, who were the primary trainers of the model.
128
129 ## Suggested Use
130
131 Use this repo as a durable artifact source and use the SQuADDS ML Space when you want
132 a stable HTTP tool surface for agents or applications.
133