Spaces:
Running
Running
Attempt to make PySR process a daemon
Browse files- gui/processing.py +67 -40
gui/processing.py
CHANGED
|
@@ -4,7 +4,6 @@ import tempfile
|
|
| 4 |
import time
|
| 5 |
from pathlib import Path
|
| 6 |
|
| 7 |
-
import numpy as np
|
| 8 |
import pandas as pd
|
| 9 |
from data import generate_data, read_csv
|
| 10 |
|
|
@@ -17,6 +16,37 @@ EMPTY_DF = lambda: pd.DataFrame(
|
|
| 17 |
)
|
| 18 |
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
def processing(
|
| 21 |
file_input,
|
| 22 |
force_run,
|
|
@@ -41,6 +71,11 @@ def processing(
|
|
| 41 |
batch_size,
|
| 42 |
):
|
| 43 |
"""Load data, then spawn a process to run the greet function."""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
if file_input is not None:
|
| 45 |
try:
|
| 46 |
X, y = read_csv(file_input, force_run)
|
|
@@ -53,31 +88,41 @@ def processing(
|
|
| 53 |
base = Path(tmpdirname)
|
| 54 |
equation_file = base / "hall_of_fame.csv"
|
| 55 |
equation_file_bkup = base / "hall_of_fame.csv.bkup"
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
X=X,
|
| 60 |
y=y,
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
|
|
|
|
|
|
| 77 |
)
|
| 78 |
-
process.start()
|
| 79 |
last_yield_time = None
|
| 80 |
-
while
|
| 81 |
if equation_file_bkup.exists():
|
| 82 |
try:
|
| 83 |
# First, copy the file to a the copy file
|
|
@@ -109,21 +154,3 @@ def processing(
|
|
| 109 |
last_yield_time = time.time()
|
| 110 |
except pd.errors.EmptyDataError:
|
| 111 |
pass
|
| 112 |
-
|
| 113 |
-
process.join()
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
def pysr_fit(
|
| 117 |
-
*,
|
| 118 |
-
X,
|
| 119 |
-
y,
|
| 120 |
-
**pysr_kwargs,
|
| 121 |
-
):
|
| 122 |
-
import pysr
|
| 123 |
-
|
| 124 |
-
model = pysr.PySRRegressor(
|
| 125 |
-
progress=False,
|
| 126 |
-
timeout_in_seconds=1000,
|
| 127 |
-
**pysr_kwargs,
|
| 128 |
-
)
|
| 129 |
-
model.fit(X, y)
|
|
|
|
| 4 |
import time
|
| 5 |
from pathlib import Path
|
| 6 |
|
|
|
|
| 7 |
import pandas as pd
|
| 8 |
from data import generate_data, read_csv
|
| 9 |
|
|
|
|
| 16 |
)
|
| 17 |
|
| 18 |
|
| 19 |
+
def pysr_fit(queue: mp.Queue, out_queue: mp.Queue):
|
| 20 |
+
import pysr
|
| 21 |
+
|
| 22 |
+
while True:
|
| 23 |
+
# Get the arguments from the queue, if available
|
| 24 |
+
args = queue.get()
|
| 25 |
+
if args is None:
|
| 26 |
+
break
|
| 27 |
+
X = args["X"]
|
| 28 |
+
y = args["y"]
|
| 29 |
+
kwargs = args["kwargs"]
|
| 30 |
+
model = pysr.PySRRegressor(
|
| 31 |
+
progress=False,
|
| 32 |
+
timeout_in_seconds=1000,
|
| 33 |
+
**kwargs,
|
| 34 |
+
)
|
| 35 |
+
model.fit(X, y)
|
| 36 |
+
out_queue.put(None)
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
class PySRProcess:
|
| 40 |
+
def __init__(self):
|
| 41 |
+
self.queue = mp.Queue()
|
| 42 |
+
self.out_queue = mp.Queue()
|
| 43 |
+
self.process = mp.Process(target=pysr_fit, args=(self.queue, self.out_queue))
|
| 44 |
+
self.process.start()
|
| 45 |
+
|
| 46 |
+
|
| 47 |
+
PERSISTENT_WRITER = None
|
| 48 |
+
|
| 49 |
+
|
| 50 |
def processing(
|
| 51 |
file_input,
|
| 52 |
force_run,
|
|
|
|
| 71 |
batch_size,
|
| 72 |
):
|
| 73 |
"""Load data, then spawn a process to run the greet function."""
|
| 74 |
+
global PERSISTENT_WRITER
|
| 75 |
+
if PERSISTENT_WRITER is None:
|
| 76 |
+
print("Starting PySR process")
|
| 77 |
+
PERSISTENT_WRITER = PySRProcess()
|
| 78 |
+
|
| 79 |
if file_input is not None:
|
| 80 |
try:
|
| 81 |
X, y = read_csv(file_input, force_run)
|
|
|
|
| 88 |
base = Path(tmpdirname)
|
| 89 |
equation_file = base / "hall_of_fame.csv"
|
| 90 |
equation_file_bkup = base / "hall_of_fame.csv.bkup"
|
| 91 |
+
# Check if queue is empty, if not, kill the process
|
| 92 |
+
# and start a new one
|
| 93 |
+
if not PERSISTENT_WRITER.queue.empty():
|
| 94 |
+
print("Restarting PySR process")
|
| 95 |
+
if PERSISTENT_WRITER.process.is_alive():
|
| 96 |
+
PERSISTENT_WRITER.process.terminate()
|
| 97 |
+
PERSISTENT_WRITER.process.join()
|
| 98 |
+
|
| 99 |
+
PERSISTENT_WRITER = PySRProcess()
|
| 100 |
+
# Write these to queue instead:
|
| 101 |
+
PERSISTENT_WRITER.queue.put(
|
| 102 |
+
dict(
|
| 103 |
X=X,
|
| 104 |
y=y,
|
| 105 |
+
kwargs=dict(
|
| 106 |
+
niterations=niterations,
|
| 107 |
+
maxsize=maxsize,
|
| 108 |
+
binary_operators=binary_operators,
|
| 109 |
+
unary_operators=unary_operators,
|
| 110 |
+
equation_file=equation_file,
|
| 111 |
+
parsimony=parsimony,
|
| 112 |
+
populations=populations,
|
| 113 |
+
population_size=population_size,
|
| 114 |
+
ncycles_per_iteration=ncycles_per_iteration,
|
| 115 |
+
elementwise_loss=elementwise_loss,
|
| 116 |
+
adaptive_parsimony_scaling=adaptive_parsimony_scaling,
|
| 117 |
+
optimizer_algorithm=optimizer_algorithm,
|
| 118 |
+
optimizer_iterations=optimizer_iterations,
|
| 119 |
+
batching=batching,
|
| 120 |
+
batch_size=batch_size,
|
| 121 |
+
),
|
| 122 |
+
)
|
| 123 |
)
|
|
|
|
| 124 |
last_yield_time = None
|
| 125 |
+
while PERSISTENT_WRITER.out_queue.empty():
|
| 126 |
if equation_file_bkup.exists():
|
| 127 |
try:
|
| 128 |
# First, copy the file to a the copy file
|
|
|
|
| 154 |
last_yield_time = time.time()
|
| 155 |
except pd.errors.EmptyDataError:
|
| 156 |
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|