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
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256 | # Copyright 2025 Patrick Armstrong
from typing import TYPE_CHECKING, ClassVar, override
import importlib
import numpy as np
from supaernova.steps.backends import AbstractModel
from supaernova.configs.steps.nflow import NFlowStepResult
from supaernova.analysis.distribution import DistributionPlotter
if TYPE_CHECKING:
from logging import Logger
from pathlib import Path
from collections.abc import Callable
from supaernova.configs.paths import PathConfig
from supaernova.configs.globals import GlobalConfig
from supaernova.steps.pae.model import PAEModelStep
from supaernova.configs.steps.nflow.model import NFlowModelConfig, NFlowStepAnalysis
from .tf import TFNFlowModel
NFlowModel = TFNFlowModel
class NFlowModelStep[Backend: str](AbstractModel[Backend]):
# --- Class Variables ---
model_backend: ClassVar[dict[str, "Callable[[], type[NFlowModel]]"]] = {
"TensorFlow": lambda: importlib.import_module(".tf", __package__).TFNFlowModel,
}
id: ClassVar[str] = "nflow_model"
def __init__(self, config: "NFlowModelConfig") -> None:
# --- Superclass Variables ---
self.options: NFlowModelConfig
self.config: GlobalConfig
self.paths: PathConfig
self.log: Logger
self.force: bool
self.verbose: bool
super().__init__(config)
# --- Config Variabls ---
self.debug: bool
self.profile: bool
self.savepath: Path
self.pae: PAEModelStep
self.results: dict[str, NFlowStepResult]
self.analysis: tuple[NFlowStepAnalysis] = self.options.analysis
@override
def _setup(self, *, pae: "PAEModelStep") -> None:
self.debug = self.options.debug
self.profile = self.options.profile
self.pae = pae
self._model(force=True)
self.savepath = self.paths.out / self.model.name
@override
def _completed(self) -> bool:
self._model(force=True)
savepath = self.savepath / self.model.ckpt_path
if not (savepath.exists() and any(savepath.iterdir())):
self.log.debug(
f"{self.name} has not completed as {savepath} does not exist"
)
return False
return True
@override
def _load(self) -> None:
self._model(force=True)
self.log.debug(f"Loading final NFlow model weights from {self.savepath}")
self.model.load_checkpoint(self.savepath)
@override
def _run(self) -> None:
self._model(force=True)
self.model.train_model(savepath=self.savepath)
@override
def _result(self) -> None:
self.log.debug(f"Saving final NFlow model weights to {self.savepath}")
self.model.save_checkpoint(self.savepath)
dt_results: dict[str, NFlowStepResult] = {}
for dt in ["train", "test"]:
data = getattr(self.model.pae.stage, f"{dt}_data")
sn_mask = getattr(self.model.pae.stage, f"{dt}_sn_mask")
spec_mask = getattr(self.model.pae.stage, f"{dt}_spec_mask")
input_phase = data.time
input_amplitude = data.amplitude
input_mask = data.mask
latents = self.model.pae(
(input_phase, input_amplitude),
training=False,
mask=input_mask * spec_mask,
)[0]
inds = np.squeeze(np.array(sn_mask).astype(np.bool_), axis=(1, 2))
latents = latents[inds]
z = latents[:, 0, :4]
if not self.model.physical_latents:
z = z[:, 1:]
log_prob = self.model(z)
u = self.model.z_to_u(z)
uz = self.model.u_to_z(u)
model_results = {
"ind": data.ind,
"sn_name": data.sn_name,
"spectra_id": data.spectra_id,
"latents": z.numpy(),
"log_prob": -log_prob.numpy(),
"z_to_u": u.numpy(),
"u_to_z": uz.numpy(),
}
dt_results[dt] = NFlowStepResult.model_validate(model_results)
self.results = dt_results
@override
def _analyse(self) -> None:
self._model()
z_labels = {}
u_labels = {}
labels = {}
ind = 0
if self.model.physical_latents:
z_labels[0] = "ΔAᵥ"
u_labels[0] = "μΔAᵥ"
labels[0] = "z/μΔAᵥ"
ind = 1
for i in range(self.model.n_u_latents):
z_labels[ind] = f"z{i}"
u_labels[ind] = f"μ{i}"
labels[ind] = f"z/μ{i}"
ind += 1
for dt in ["train", "test"]:
results = self.results[dt]
gaussian = np.random.normal(0, 1, (results.z_to_u.size**2, ind))
if self.analysis.plot_u_latents is not None:
if not isinstance(self.analysis.plot_u_latents, list):
self.analysis.plot_u_latents = [self.analysis.plot_u_latents]
for opts in self.analysis.plot_u_latents:
o = opts.model_copy()
if o.labels is None:
o.labels = {"gaussian": u_labels, "u_latents": u_labels}
if o.name is None:
o.name = "u_latents"
if o.savepath is None:
o.savepath = self.paths.plots / dt / str(self.model.seed)
o.savepath.mkdir(parents=True, exist_ok=True)
DistributionPlotter.plot_corner(
{"gaussian": gaussian, "u_latents": results.z_to_u},
o,
statistics="max_central",
shade_alpha=0.0,
plot_cloud=True,
)
if self.analysis.plot_z_latents is not None:
if not isinstance(self.analysis.plot_z_latents, list):
self.analysis.plot_z_latents = [self.analysis.plot_z_latents]
for opts in self.analysis.plot_z_latents:
o = opts.model_copy()
if o.labels is None:
o.labels = z_labels
if o.name is None:
o.name = "z_latents"
if o.savepath is None:
o.savepath = self.paths.plots / dt / str(self.model.seed)
o.savepath.mkdir(parents=True, exist_ok=True)
DistributionPlotter.plot_corner(
results.u_to_z,
o,
statistics="max_central",
shade_alpha=0.0,
plot_cloud=True,
)
if self.analysis.plot_latents is not None:
if not isinstance(self.analysis.plot_latents, list):
self.analysis.plot_latents = [self.analysis.plot_latents]
for opts in self.analysis.plot_latents:
o = opts.model_copy()
if o.labels is None:
o.labels = {
"z_latents": labels,
"u_latents": labels,
}
if o.name is None:
o.name = "latents"
if o.savepath is None:
o.savepath = self.paths.plots / dt / str(self.model.seed)
o.savepath.mkdir(parents=True, exist_ok=True)
u_latents = self.model.z_to_u(results.latents, permute=True).numpy()
z_latents = self.model.u_to_z(u_latents, permute=True).numpy()
DistributionPlotter.plot_corner(
{"u_latents": u_latents, "z_latents": z_latents},
o,
statistics="max_central",
shade_alpha=0.0,
plot_cloud=True,
)
if self.analysis.plot_latent_steps is not None:
if not isinstance(self.analysis.plot_latent_steps, list):
self.analysis.plot_latent_steps = [self.analysis.plot_latent_steps]
for opts in self.analysis.plot_latent_steps:
num_steps = len(self.model.flow.bijector.bijectors) + 1
for step in range(num_steps):
step_latents, is_shift = self.model.z_to_u_steps(
results.latents, step, permute=True
)
if is_shift:
continue
o = opts.model_copy()
if o.labels is None:
o.labels = {
"gaussian": labels,
f"step_{step}_latents": labels,
}
if o.name is None:
o.name = f"step_{step}_latent_steps"
if o.savepath is None:
o.savepath = (
self.paths.plots / dt / str(self.model.seed) / "steps"
)
o.savepath.mkdir(parents=True, exist_ok=True)
DistributionPlotter.plot_corner(
{
"gaussian": gaussian,
f"step_{step}_latents": step_latents.numpy(),
},
o,
statistics="max_central",
shade_alpha=0.0,
plot_cloud=True,
)
#
# === NFlowModel Specific Functions ===
#
|