File size: 9,545 Bytes
9fe46c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from mace.calculators import mace_mp
from ase.io import read, write
from ase.io.trajectory import Trajectory
from ase.optimize import FIRE, LBFGS
from matplotlib.patches import Patch
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
import time

# Load POSCAR structure
atoms = read('POSCAR', format='vasp')
atoms.set_pbc(True)
atoms.set_positions(atoms.get_positions().astype(np.float64))

# Load custom-trained MACE model
atoms.calc = mace_mp(
    model="6_MACE.model",
    default_dtype="float64",
    device="cuda:1",
)

# Verify calculator works
print("Initial energy:", atoms.get_potential_energy(), "eV")
print("Initial max force:", np.max(np.abs(atoms.get_forces())), "eV/Å")

# --- Loggers ---
steps      = []
energies   = []
fmax_vals  = []
stages     = []
timestamps = []

t_run_start = time.perf_counter()

class EnergyLogger:
    def __init__(self, atoms, stage_label):
        self.atoms       = atoms
        self.stage_label = stage_label
        self.step        = 0

    def __call__(self):
        e = self.atoms.get_potential_energy()
        f = np.max(np.abs(self.atoms.get_forces()))
        steps.append(len(steps))
        energies.append(e)
        fmax_vals.append(f)
        stages.append(self.stage_label)
        timestamps.append(time.perf_counter() - t_run_start)
        self.step += 1

class UnifiedLogger:
    def __init__(self, atoms, stage_label, logfile):
        self.atoms         = atoms
        self.stage_label   = stage_label
        self.logfile       = logfile
        self.step          = 0
        self.t_stage_start = None

    def __call__(self):
        if self.t_stage_start is None:
            self.t_stage_start = time.perf_counter()

        e       = self.atoms.get_potential_energy()
        f       = np.max(np.abs(self.atoms.get_forces()))
        elapsed = time.perf_counter() - self.t_stage_start

        with open(self.logfile, 'a') as fh:
            if self.step == 0:
                fh.write(f"\n# --- {self.stage_label} ---\n")
                fh.write(f"{'Step':>6}  {'Energy (eV)':>14}  {'fmax (eV/Å)':>12}  {'Time (s)':>10}\n")
            fh.write(f"{self.step:>6}  {e:>14.6f}  {f:>12.6f}  {elapsed:>10.2f}\n")
        self.step += 1

LOG_FILE  = 'relax_MC.log'
TRAJ_FILE = 'relax_MC.traj'

open(LOG_FILE, 'w').close()
traj = Trajectory(TRAJ_FILE, mode='w', atoms=atoms)

# ----------------------------------------------------------
# Stage 1: FIRE
# ----------------------------------------------------------
fire_energy_logger  = EnergyLogger(atoms, stage_label="FIRE")
fire_unified_logger = UnifiedLogger(atoms, stage_label="FIRE", logfile=LOG_FILE)

fire = FIRE(atoms, logfile=None, dt=0.05, maxstep=0.05, dtmax=0.5, Nmin=10, finc=1.05, fdec=0.5)
fire.attach(fire_energy_logger,  interval=1)
fire.attach(fire_unified_logger, interval=1)
fire.attach(traj.write,          interval=1)

t_fire_start = time.perf_counter()
fire.run(fmax=0.1)
t_fire = time.perf_counter() - t_fire_start

nsteps_fire = fire.get_number_of_steps()
print(f"✓ FIRE done : {nsteps_fire} steps | "
      f"fmax={np.max(np.abs(atoms.get_forces())):.4f} eV/Å | "
      f"time={t_fire:.1f}s ({t_fire/60:.2f} min)")

# ----------------------------------------------------------
# Stage 2: LBFGS
# ----------------------------------------------------------
lbfgs_energy_logger  = EnergyLogger(atoms, stage_label="LBFGS")
lbfgs_unified_logger = UnifiedLogger(atoms, stage_label="LBFGS", logfile=LOG_FILE)

lbfgs = LBFGS(atoms, logfile=None, maxstep=0.05, memory=100)
lbfgs.attach(lbfgs_energy_logger,  interval=1)
lbfgs.attach(lbfgs_unified_logger, interval=1)
lbfgs.attach(traj.write,           interval=1)

t_lbfgs_start = time.perf_counter()
lbfgs.run(fmax=0.01)
t_lbfgs = time.perf_counter() - t_lbfgs_start

traj.close()

nsteps_lbfgs = lbfgs.get_number_of_steps()
print(f"✓ LBFGS done: {nsteps_lbfgs} steps | "
      f"fmax={np.max(np.abs(atoms.get_forces())):.4f} eV/Å | "
      f"time={t_lbfgs:.1f}s ({t_lbfgs/60:.2f} min)")

t_total      = t_fire + t_lbfgs
final_energy = atoms.get_potential_energy()

# --- Masks and timing arrays ---
stages_arr = np.array(stages)
times_arr  = np.array(timestamps)
steps_arr  = np.array(steps)
fire_mask  = stages_arr == "FIRE"
lbfgs_mask = stages_arr == "LBFGS"

x_fire  = times_arr[fire_mask]
x_lbfgs = times_arr[lbfgs_mask]
s_fire  = steps_arr[fire_mask]
s_lbfgs = steps_arr[lbfgs_mask]

# Per-step dt
dt_fire  = np.diff(x_fire,  prepend=x_fire[0])  if len(x_fire)  > 0 else np.array([])
dt_lbfgs = np.diff(x_lbfgs, prepend=x_lbfgs[0]) if len(x_lbfgs) > 0 else np.array([])

avg_fire  = dt_fire.mean()  if len(dt_fire)  > 0 else 0.0
avg_lbfgs = dt_lbfgs.mean() if len(dt_lbfgs) > 0 else 0.0

transition_step = s_lbfgs[0] if any(lbfgs_mask) else None

# Combined for bar chart
all_steps_arr = np.concatenate([s_fire, s_lbfgs])
all_dt        = np.concatenate([dt_fire, dt_lbfgs])
all_colors    = ['#E07B39'] * len(s_fire) + ['#1D9E75'] * len(s_lbfgs)
avg_total     = all_dt.mean() if len(all_dt) > 0 else 0.0

# --- Helper ---
def fmt_time(seconds):
    s = int(seconds)
    h, rem = divmod(s, 3600)
    m, sec = divmod(rem, 60)
    if h > 0:
        return f"{h}h {m}m {sec}s"
    elif m > 0:
        return f"{m}m {sec}s"
    else:
        return f"{sec}s"

# --- Console summary ---
print(f"\n✓ Done! Final energy : {final_energy:.4f} eV")
print(f"Total steps          : {nsteps_fire + nsteps_lbfgs} (FIRE={nsteps_fire}, LBFGS={nsteps_lbfgs})")
print(f"Final fmax           : {np.max(np.abs(atoms.get_forces())):.4f} eV/Å")
print(f"FIRE  wall time      : {fmt_time(t_fire)}  avg {avg_fire:.2f}s/step  "
      f"min {dt_fire.min():.2f}s  max {dt_fire.max():.2f}s")
print(f"LBFGS wall time      : {fmt_time(t_lbfgs)}  avg {avg_lbfgs:.2f}s/step  "
      f"min {dt_lbfgs.min():.2f}s  max {dt_lbfgs.max():.2f}s")
print(f"Total wall time      : {fmt_time(t_total)}")

# Save structures
write('POSCAR-relaxed',  atoms, format='vasp', vasp5=True, sort=True)
write('CONTCAR-relaxed', atoms, format='vasp', direct=True)

# --- Timing summary to log ---
with open(LOG_FILE, 'a') as fh:
    fh.write(f"\n# --- Timing Summary ---\n")
    fh.write(f"{'Stage':<8}  {'Steps':>6}  {'Total':>12}  {'Avg (s/step)':>13}  "
             f"{'Min (s)':>8}  {'Max (s)':>8}\n")
    fh.write(f"{'FIRE':<8}  {nsteps_fire:>6}  {fmt_time(t_fire):>12}  {avg_fire:>13.3f}  "
             f"{dt_fire.min():>8.3f}  {dt_fire.max():>8.3f}\n")
    fh.write(f"{'LBFGS':<8}  {nsteps_lbfgs:>6}  {fmt_time(t_lbfgs):>12}  {avg_lbfgs:>13.3f}  "
             f"{dt_lbfgs.min():>8.3f}  {dt_lbfgs.max():>8.3f}\n")
    fh.write(f"{'Total':<8}  {nsteps_fire+nsteps_lbfgs:>6}  {fmt_time(t_total):>12}  "
             f"{'—':>13}  {'—':>8}  {'—':>8}\n")

# --- Plot ---
fig, axes = plt.subplots(3, 1, figsize=(8, 8), sharex=True)
fig.suptitle('MACE Relaxation — FIRE → LBFGS', fontsize=13, fontweight='normal')
ax1, ax2, ax3 = axes

# --- Energy panel ---
ax1.plot(s_fire,  np.array(energies)[fire_mask],
         color='#E07B39', linewidth=1.5, marker='o', markersize=2.5,
         markeredgewidth=0, label='FIRE')
ax1.plot(s_lbfgs, np.array(energies)[lbfgs_mask],
         color='#1D9E75', linewidth=1.5, marker='o', markersize=2.5,
         markeredgewidth=0, label='LBFGS')
ax1.axhline(y=final_energy, color='#D85A30', linestyle='--',
            linewidth=1, label=f'Final: {final_energy:.4f} eV')
if transition_step is not None:
    ax1.axvline(x=transition_step, color='gray', linestyle=':', linewidth=1, label='FIRE→LBFGS')
ax1.set_ylabel('Energy (eV)', fontsize=11)
ax1.legend(fontsize=9, framealpha=0.7)
ax1.yaxis.set_major_formatter(ticker.FormatStrFormatter('%.3f'))
ax1.grid(True, alpha=0.3, linestyle='--')
ax1.tick_params(labelsize=9)

# --- fmax panel ---
ax2.plot(s_fire,  np.array(fmax_vals)[fire_mask],
         color='#E07B39', linewidth=1.5, marker='o', markersize=2.5,
         markeredgewidth=0, label='FIRE')
ax2.plot(s_lbfgs, np.array(fmax_vals)[lbfgs_mask],
         color='#1D9E75', linewidth=1.5, marker='o', markersize=2.5,
         markeredgewidth=0, label='LBFGS')
ax2.axhline(y=0.01, color='#D85A30', linestyle='--',
            linewidth=1, label='Convergence (0.01 eV/Å)')
ax2.axhline(y=0.1,  color='gray',    linestyle=':',
            linewidth=1, label='FIRE target (0.1 eV/Å)')
if transition_step is not None:
    ax2.axvline(x=transition_step, color='gray', linestyle=':', linewidth=1)
ax2.set_ylabel('Max Force (eV/Å)', fontsize=11)
ax2.legend(fontsize=9, framealpha=0.7)
ax2.set_yscale('log')
ax2.grid(True, alpha=0.3, linestyle='--')
ax2.tick_params(labelsize=9)

# --- Wall time per step bar chart ---
ax3.bar(all_steps_arr, all_dt, color=all_colors, alpha=0.75, width=0.6)
ax3.axhline(y=avg_total, color='#D85A30', linestyle='--', linewidth=1,
            label=f'Avg: {avg_total:.3f}s  |  Total: {fmt_time(t_total)}')
if transition_step is not None:
    ax3.axvline(x=transition_step, color='gray', linestyle=':', linewidth=1)

legend_handles = [
    plt.Line2D([0], [0], color='#D85A30', linestyle='--',
               linewidth=1, label=f'Avg: {avg_total:.3f}s  |  Total: {fmt_time(t_total)}'),
]
ax3.legend(handles=legend_handles, fontsize=9, framealpha=0.7)
ax3.set_ylabel('Wall Time (s)', fontsize=11)
ax3.set_xlabel('Ionic Step', fontsize=11)
ax3.grid(True, alpha=0.3, linestyle='--', axis='y')
ax3.tick_params(labelsize=9)

plt.tight_layout()
plt.savefig('MC_relaxation_plot_FIRE_LBFGS.png', dpi=150, bbox_inches='tight')
plt.show()
print("Plot saved as MC_relaxation_plot_FIRE_LBFGS.png")