Merge pull request #119 from MiraGeoscience/GEOPY-2466

GEOPY-2466: Parallelize the Sweeps

Author: domfournier

simpeg/dask/electromagnetics/frequency_domain/simulation.py

@@ -7,7 +7,6 @@
 import scipy.sparse as sp
 
 from dask import array, compute, delayed
-from dask.distributed import get_client
 from simpeg.dask.utils import get_parallel_blocks
 from simpeg.electromagnetics.natural_source.sources import PlanewaveXYPrimary
 import zarr
@@ -104,20 +103,17 @@ def getSourceTerm(self, freq, source=None):
 
     if source is None:
 
-        try:
-            client = get_client()
-            sim = client.scatter(self, workers=self.worker)
-        except ValueError:
-            client = None
-            sim = self
-
+        client, worker = self._get_client_worker()
         source_list = self.survey.get_sources_by_frequency(freq)
         source_blocks = np.array_split(
-            np.arange(len(source_list)), self.n_threads(client=client)
+            np.arange(len(source_list)), self.n_threads(client=client, worker=worker)
         )
 
         if client:
-            source_list = client.scatter(source_list, workers=self.worker)
+            sim = client.scatter(self, workers=self.worker)
+            source_list = client.scatter(source_list, workers=worker)
+        else:
+            sim = self
 
         block_compute = []
 
@@ -127,9 +123,7 @@ def getSourceTerm(self, freq, source=None):
 
             if client:
                 block_compute.append(
-                    client.submit(
-                        source_eval, sim, source_list, block, workers=self.worker
-                    )
+                    client.submit(source_eval, sim, source_list, block, workers=worker)
                 )
             else:
                 block_compute.append(source_eval(sim, source_list, block))
@@ -221,12 +215,7 @@ def compute_J(self, m, f=None):
     fields_array = f[:, self._solutionType]
     blocks_receiver_derivs = []
 
-    try:
-        client = get_client()
-        worker = self.worker
-    except ValueError:
-        client = None
-        worker = None
+    client, worker = self._get_client_worker()
 
     if client:
         fields_array = client.scatter(f[:, self._solutionType], workers=worker)

simpeg/dask/electromagnetics/static/resistivity/simulation.py

@@ -3,7 +3,7 @@
 from ....simulation import getJtJdiag, Jvec, Jtvec, Jmatrix
 
 from .....utils import Zero
-from dask.distributed import get_client
+
 import dask.array as da
 import numpy as np
 from scipy import sparse as sp
@@ -163,22 +163,23 @@ def getSourceTerm(self):
             source_list = self.survey.source_list
 
         indices = np.arange(len(source_list))
-        try:
 
-            client = get_client()
-            sim = client.scatter(self, workers=self.worker)
-            future_list = client.scatter(source_list, workers=self.worker)
-            indices = np.array_split(indices, self.n_threads(client=client))
+        client, worker = self._get_client_worker()
+
+        if client:
+            sim = client.scatter(self, workers=worker)
+            future_list = client.scatter(source_list, workers=worker)
+            indices = np.array_split(
+                indices, self.n_threads(client=client, worker=worker)
+            )
             blocks = []
             for ind in indices:
                 blocks.append(
-                    client.submit(
-                        source_eval, sim, future_list, ind, workers=self.worker
-                    )
+                    client.submit(source_eval, sim, future_list, ind, workers=worker)
                 )
 
             blocks = sp.hstack(client.gather(blocks))
-        except ValueError:
+        else:
             blocks = source_eval(self, source_list, indices)
 
         self._q = blocks

simpeg/dask/electromagnetics/time_domain/simulation.py

@@ -9,7 +9,7 @@
 import numpy as np
 import scipy.sparse as sp
 from dask import array, delayed
-from dask.distributed import get_client
+
 
 from time import time
 from simpeg.dask.utils import get_parallel_blocks
@@ -89,10 +89,7 @@ def compute_J(self, m, f=None):
     if f is None:
         f, Ainv = self.fields(m=m, return_Ainv=True)
 
-    try:
-        client = get_client()
-    except ValueError:
-        client = None
+    client, worker = self._get_client_worker()
 
     ftype = self._fieldType + "Solution"
     sens_name = self.sensitivity_path[:-5]
@@ -118,22 +115,22 @@ def compute_J(self, m, f=None):
     blocks = get_parallel_blocks(
         self.survey.source_list,
         compute_row_size,
-        thread_count=self.n_threads(client=client),
+        thread_count=self.n_threads(client=client, worker=worker),
     )
     fields_array = f[:, ftype, :]
 
     if len(self.survey.source_list) == 1:
         fields_array = fields_array[:, np.newaxis, :]
 
     times_field_derivs, Jmatrix = compute_field_derivs(
-        self, f, blocks, Jmatrix, fields_array.shape, client
+        self, f, blocks, Jmatrix, fields_array.shape
     )
 
     ATinv_df_duT_v = [[] for _ in blocks]
 
     if client:
-        fields_array = client.scatter(fields_array, workers=self.worker)
-        sim = client.scatter(self, workers=self.worker)
+        fields_array = client.scatter(fields_array, workers=worker)
+        sim = client.scatter(self, workers=worker)
     else:
         delayed_compute_rows = delayed(compute_rows)
         sim = self
@@ -161,7 +158,7 @@ def compute_J(self, m, f=None):
             )
 
         if client:
-            field_derivatives = client.scatter(ATinv_df_duT_v, workers=self.worker)
+            field_derivatives = client.scatter(ATinv_df_duT_v, workers=worker)
         else:
             field_derivatives = ATinv_df_duT_v
 
@@ -181,7 +178,7 @@ def compute_J(self, m, f=None):
                         field_derivatives,
                         fields_array,
                         time_mask,
-                        workers=self.worker,
+                        workers=worker,
                     )
                 )
             else:
@@ -267,17 +264,19 @@ def evaluate_receivers(block, mesh, time_mesh, fields, fields_array):
     return np.hstack(data)
 
 
-def compute_field_derivs(self, fields, blocks, Jmatrix, fields_shape, client):
+def compute_field_derivs(self, fields, blocks, Jmatrix, fields_shape):
     """
     Compute the derivative of the fields
     """
     delayed_chunks = []
 
+    client, worker = self._get_client_worker()
+
     if client:
-        mesh = client.scatter(self.mesh, workers=self.worker)
-        time_mesh = client.scatter(self.time_mesh, workers=self.worker)
-        fields = client.scatter(fields, workers=self.worker)
-        source_list = client.scatter(self.survey.source_list, workers=self.worker)
+        mesh = client.scatter(self.mesh, workers=worker)
+        time_mesh = client.scatter(self.time_mesh, workers=worker)
+        fields = client.scatter(fields, workers=worker)
+        source_list = client.scatter(self.survey.source_list, workers=worker)
     else:
         mesh = self.mesh
         time_mesh = self.time_mesh
@@ -300,7 +299,7 @@ def compute_field_derivs(self, fields, blocks, Jmatrix, fields_shape, client):
                     time_mesh,
                     fields,
                     self.model.size,
-                    workers=self.worker,
+                    workers=worker,
                 )
             )
         else:
@@ -537,24 +536,22 @@ def dpred(self, m=None, f=None):
             "simulation.survey = survey"
         )
 
-    try:
-        client = get_client()
-    except ValueError:
-        client = None
+    client, worker = self._get_client_worker()
 
     if f is None:
         f = self.fields(m)
 
     delayed_chunks = []
 
     source_block = np.array_split(
-        np.arange(len(self.survey.source_list)), self.n_threads(client=client)
+        np.arange(len(self.survey.source_list)),
+        self.n_threads(client=client, worker=worker),
     )
     if client:
-        mesh = client.scatter(self.mesh, workers=self.worker)
-        time_mesh = client.scatter(self.time_mesh, workers=self.worker)
-        fields = client.scatter(f, workers=self.worker)
-        source_list = client.scatter(self.survey.source_list, workers=self.worker)
+        mesh = client.scatter(self.mesh, workers=worker)
+        time_mesh = client.scatter(self.time_mesh, workers=worker)
+        fields = client.scatter(f, workers=worker)
+        source_list = client.scatter(self.survey.source_list, workers=worker)
     else:
         mesh = self.mesh
         time_mesh = self.time_mesh
@@ -575,7 +572,7 @@ def dpred(self, m=None, f=None):
                     mesh,
                     time_mesh,
                     fields,
-                    workers=self.worker,
+                    workers=worker,
                 )
             )
         else:

simpeg/dask/objective_function.py

@@ -7,7 +7,7 @@
 from typing import Callable
 import numpy as np
 
-from dask.distributed import Client, Future
+from dask.distributed import Client, Future, get_client
 from ..data_misfit import L2DataMisfit
 
 from simpeg.utils import validate_list_of_types
@@ -145,9 +145,7 @@ def _validate_type_or_future_of_type(
 ):
 
     if workers is None:
-        workers = [
-            (worker.worker_address,) for worker in client.cluster.workers.values()
-        ]
+        workers = [(worker,) for worker in client.nthreads()]
 
     objects = validate_list_of_types(
         property_name, objects, obj_type, ensure_unique=True
@@ -262,6 +260,9 @@ def client(self):
 
     @client.setter
     def client(self, client):
+        if client is None:
+            client = get_client()
+
         if not isinstance(client, Client):
             raise TypeError("client must be a dask.distributed.Client")
 
@@ -279,6 +280,23 @@ def workers(self, workers):
         if not isinstance(workers, list | type(None)):
             raise TypeError("workers must be a list of strings")
 
+        available_workers = [(worker,) for worker in self.client.nthreads()]
+
+        if workers is None:
+            workers = available_workers
+
+        if not isinstance(workers, list) or not all(
+            isinstance(w, tuple) for w in workers
+        ):
+            raise TypeError("Workers must be a list of tuple[str].")
+
+        invalid_workers = [w for w in workers if w not in available_workers]
+        if invalid_workers:
+            raise ValueError(
+                f"The following workers are not available: {invalid_workers}. "
+                f"Available workers are: {available_workers}."
+            )
+
         self._workers = workers
 
     def deriv(self, m, f=None):

simpeg/dask/potential_fields/base.py

@@ -1,7 +1,7 @@
 import numpy as np
 
 from ...potential_fields.base import BasePFSimulation as Sim
-from dask.distributed import get_client
+
 import os
 from dask import delayed, array, config
 from dask.diagnostics import ProgressBar
@@ -60,13 +60,10 @@ def linear_operator(self):
     )
     block_split = np.array_split(self.survey.receiver_locations, n_blocks)
 
-    try:
-        client = get_client()
-    except ValueError:
-        client = None
+    client, worker = self._get_client_worker()
 
     if client:
-        sim = client.scatter(self, workers=self.worker)
+        sim = client.scatter(self, workers=worker)
     else:
         delayed_compute = delayed(block_compute)
 
@@ -79,7 +76,7 @@ def linear_operator(self):
                     sim,
                     block,
                     self.survey.components,
-                    workers=self.worker,
+                    workers=worker,
                 )
             )
         else: