Aleksei Ustimenko commited on
Commit
6f01cc0
·
1 Parent(s): 4e645a1

Publish HamiltonZero 0.1.1 package

Browse files
Files changed (2) hide show
  1. README.md +18 -10
  2. pyproject.toml +22 -1
README.md CHANGED
@@ -14,12 +14,21 @@ wavefunctions of quantum spin Hamiltonians. It exposes three workflows:
14
 
15
  ## Installation
16
 
17
- HamiltonZero requires Python 3.12 and JAX-compatible accelerator drivers.
 
18
 
19
  ```bash
20
- python -m pip install .
21
  ```
22
 
 
 
 
 
 
 
 
 
23
  The package pins `jax==0.11.0` and `jaxlib==0.11.0`. Install the accelerator
24
  plugin appropriate for the host using the standard JAX instructions.
25
 
@@ -71,10 +80,9 @@ does this from the input system automatically.
71
 
72
  The public API follows the textbook convention
73
 
74
- \[
75
- H = \sum_{i<j} S_i^T J_{ij} S_j + \sum_i h_i^T S_i,
76
- \qquad S=\sigma/2.
77
- \]
78
 
79
  Construct and save a system from a simple undirected NetworkX graph:
80
 
@@ -153,8 +161,8 @@ There is no additional bit reversal: applying one would corrupt the mapping.
153
  Both arrays include padded virtual leaves when the model width exceeds the
154
  physical site count. A complete runnable version that prints sample means and
155
  standard deviations is in
156
- [`examples/compiled_inference.py`](examples/compiled_inference.py).
157
- [`examples/j1j2_4x4_route.ipynb`](examples/j1j2_4x4_route.ipynb) constructs a
158
  periodic 4-by-4 J1-J2 model from NetworkX and visualizes the returned order as
159
  the successive cells of the compiled binary merge tree. Install its plotting
160
  dependencies with `python -m pip install '.[notebooks]'`.
@@ -206,7 +214,7 @@ public textbook units above.
206
  spins.
207
 
208
  Large-N files store physical sites only; the loader reconstructs power-of-two
209
- padding in memory. See [`datasets/README.md`](datasets/README.md) for the full
210
  inventory and sparse exchange encoding.
211
 
212
  ## Train
@@ -303,4 +311,4 @@ licensed under Apache-2.0, copyright Simulacra Research Inc. The vendored
303
  KFAC-JAX fork and JAX-derived large-N attention kernel remain under
304
  Apache-2.0. The Microsoft-Folx-derived attention forward and reverse-mode
305
  kernels remain under MIT. See
306
- [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md).
 
14
 
15
  ## Installation
16
 
17
+ HamiltonZero requires Python 3.12 or newer and JAX-compatible accelerator
18
+ drivers.
19
 
20
  ```bash
21
+ python -m pip install hamiltonzero
22
  ```
23
 
24
+ To install from a source checkout instead, run `python -m pip install .` in
25
+ the repository root.
26
+
27
+ Example configurations and research datasets are repository assets rather
28
+ than wheel data. Clone the matching
29
+ [`v0.1.1` source tree](https://github.com/simulacra-research/HamiltonZero/tree/v0.1.1)
30
+ to run the documented commands unchanged.
31
+
32
  The package pins `jax==0.11.0` and `jaxlib==0.11.0`. Install the accelerator
33
  plugin appropriate for the host using the standard JAX instructions.
34
 
 
80
 
81
  The public API follows the textbook convention
82
 
83
+ ```text
84
+ H = sum_(i<j) S_i^T J_ij S_j + sum_i h_i^T S_i, S = sigma/2.
85
+ ```
 
86
 
87
  Construct and save a system from a simple undirected NetworkX graph:
88
 
 
161
  Both arrays include padded virtual leaves when the model width exceeds the
162
  physical site count. A complete runnable version that prints sample means and
163
  standard deviations is in
164
+ [`examples/compiled_inference.py`](https://github.com/simulacra-research/HamiltonZero/blob/v0.1.1/examples/compiled_inference.py).
165
+ [`examples/j1j2_4x4_route.ipynb`](https://github.com/simulacra-research/HamiltonZero/blob/v0.1.1/examples/j1j2_4x4_route.ipynb) constructs a
166
  periodic 4-by-4 J1-J2 model from NetworkX and visualizes the returned order as
167
  the successive cells of the compiled binary merge tree. Install its plotting
168
  dependencies with `python -m pip install '.[notebooks]'`.
 
214
  spins.
215
 
216
  Large-N files store physical sites only; the loader reconstructs power-of-two
217
+ padding in memory. See [`datasets/README.md`](https://github.com/simulacra-research/HamiltonZero/blob/v0.1.1/datasets/README.md) for the full
218
  inventory and sparse exchange encoding.
219
 
220
  ## Train
 
311
  KFAC-JAX fork and JAX-derived large-N attention kernel remain under
312
  Apache-2.0. The Microsoft-Folx-derived attention forward and reverse-mode
313
  kernels remain under MIT. See
314
+ [`THIRD_PARTY_NOTICES.md`](https://github.com/simulacra-research/HamiltonZero/blob/v0.1.1/THIRD_PARTY_NOTICES.md).
pyproject.toml CHANGED
@@ -4,10 +4,25 @@ build-backend = "setuptools.build_meta"
4
 
5
  [project]
6
  name = "hamiltonzero"
7
- version = "0.1.0"
8
  description = "Compiled neural wavefunctions for quantum spin systems"
9
  readme = "README.md"
10
  requires-python = ">=3.12"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  license = "Apache-2.0 AND MIT"
12
  license-files = [
13
  "LICENSE",
@@ -33,6 +48,12 @@ dependencies = [
33
  "typing-extensions>=4.15",
34
  ]
35
 
 
 
 
 
 
 
36
  [project.optional-dependencies]
37
  notebooks = [
38
  "jupyterlab>=4.4",
 
4
 
5
  [project]
6
  name = "hamiltonzero"
7
+ version = "0.1.1"
8
  description = "Compiled neural wavefunctions for quantum spin systems"
9
  readme = "README.md"
10
  requires-python = ">=3.12"
11
+ authors = [
12
+ {name = "Simulacra Research Inc."},
13
+ ]
14
+ keywords = [
15
+ "neural quantum states",
16
+ "quantum many-body physics",
17
+ "quantum Monte Carlo",
18
+ ]
19
+ classifiers = [
20
+ "Development Status :: 3 - Alpha",
21
+ "Intended Audience :: Science/Research",
22
+ "Programming Language :: Python :: 3",
23
+ "Programming Language :: Python :: 3.12",
24
+ "Topic :: Scientific/Engineering :: Physics",
25
+ ]
26
  license = "Apache-2.0 AND MIT"
27
  license-files = [
28
  "LICENSE",
 
48
  "typing-extensions>=4.15",
49
  ]
50
 
51
+ [project.urls]
52
+ Homepage = "https://github.com/simulacra-research/HamiltonZero"
53
+ Repository = "https://github.com/simulacra-research/HamiltonZero"
54
+ Issues = "https://github.com/simulacra-research/HamiltonZero/issues"
55
+ "Model weights" = "https://huggingface.co/simulacra-research/HamiltonZero"
56
+
57
  [project.optional-dependencies]
58
  notebooks = [
59
  "jupyterlab>=4.4",