sync

spacy/cli/ray_param_server.py

@@ -1,23 +1,44 @@
 """Parameter Server distributed training with Ray."""
-
+import threading
 import ray
 from wasabi import msg
 from .. import util
+from spacy.cli.ray_utils import create_optimizer
 
 class OptimizerWorker:
-    def __init__(self, config_path, world_size, sync=True):
-        self.optimizer = _create_optimizer(config_path)
+    def __init__(self, config_path, world_size):
+        self.optimizer = create_optimizer(config_path)
+        self.new_weights = None
+        self.barrier = threading.Barrier(world_size)
+        self.lock = threading.Lock()
+        self.waiting = 0
         self.weights_dict = {}
+        self.grad_dict = {}
         self.world_size = world_size
-        self.sync = sync
 
-    def call(self, rank, key, weights, gradient, *, lr_scale=1.0):
-        if key not in self.weights_dict:
-            self.weights_dict[key] = weights.copy()
-        new_weights, new_grads = self.optimizer(
-            key, self.weights_dict[key], gradient.copy(), lr_scale=lr_scale)
-        self.weights_dict[key] = new_weights
-        return new_weights, new_grads
+    def call(self, key, weights, gradient, *, lr_scale=1.0):
+        self.lock.acquire()
+
+        if self.waiting < self.world_size - 1:
+            if self.waiting == 0:
+                self.gradient[key] = gradient.copy()
+                self.weights_dict[key] = weights.copy()
+            else:
+                self.gradient[key] += gradient
+            self.waiting = self.barrier.n_waiting + 1
+            self.lock.release()
+            self.barrier.wait()
+        else:
+            self.gradient[key] += gradient
+            self.lock.release()
+            self.gradient[key] /= self.world_size
+            new_weights, new_grads = self.optimizer(
+                key, self.weights_dict[key], self.gradient[key], lr_scale=lr_scale)
+            self.weights_dict[key] = new_weights
+            self.gradient[key] = new_grads
+            self.waiting = 0
+            self.barrier.wait()
+        return self.weights_dict[key], self.gradient[key]
 
     def fetch(self):
         return self.optimizer
@@ -40,7 +61,7 @@ class RayOptimizer:
         self.local_optimizer = ray.get(self.optimizer.fetch.remote())
 
     def __call__(self, *args, **kwargs):
-        weights, grads = ray.get(self.optimizer.call.remote(self.rank, *args, **kwargs))
+        weights, grads = ray.get(self.optimizer.call.remote(*args, **kwargs))
         return weights.copy(), grads.copy()
 
     def __getattr__(self, name):

spacy/cli/ray_utils.py

@@ -10,7 +10,7 @@ nccl = None
 from typing import Dict, Optional, Union, Tuple, List, cast
 from thinc.types import FloatsXd
 
-def _create_optimizer(config_path):
+def create_optimizer(config_path):
     msg.info(f"Loading config from: {config_path}")
     config = util.load_config(config_path, create_objects=False)
     util.fix_random_seed(config["training"]["seed"])  # Fix this.
@@ -41,7 +41,7 @@ class AllreduceOptimizer:
         import cupy as cp
         global nccl
         from cupy.cuda import nccl
-        self.optimizer = _create_optimizer(config_path)
+        self.optimizer = create_optimizer(config_path)
         self.communicator = communicator
         self.weights_synced = set()