Apply review comments

fl4health/clients/fenda_client.py

@@ -142,10 +142,10 @@ def get_perFCL_loss(
     ) -> Tuple[torch.Tensor, torch.Tensor]:
         """
         PerFCL loss consists of two contrastive losses.
-        First one airms to enhance the similarity between the current global features and aggregated global feature
+        First one aims to enhance the similarity between the current global features and aggregated global features
         as positive pairs while reducing the similarity between the current global features and old global
         features as negative pairs.
-        Second one airms to enhance the similarity between the current local features and old local feature
+        Second one aims to enhance the similarity between the current local features and old local features
         as positive pairs while reducing the similarity between the current local features and aggregated global
         features as negative pairs.
         """

research/flamby/fed_heart_disease/fenda/client.py

@@ -33,7 +33,7 @@ def __init__(
         loss_meter_type: LossMeterType = LossMeterType.AVERAGE,
         metric_meter_type: MetricMeterType = MetricMeterType.ACCUMULATION,
         checkpointer: Optional[TorchCheckpointer] = None,
-        type_run: str = "cos_sim",
+        type_run: str = "vanilla",
     ) -> None:
         super().__init__(
             data_path=data_path,

research/flamby/fed_heart_disease/fenda/server.py

@@ -20,7 +20,7 @@
 )
 
 
-def main(config: Dict[str, Any], server_address: str, run_name: str) -> None:
+def main(config: Dict[str, Any], server_address: str) -> None:
     # This function will be used to produce a config that is sent to each client to initialize their own environment
     fit_config_fn = partial(
         fit_config,
@@ -77,14 +77,8 @@ def main(config: Dict[str, Any], server_address: str, run_name: str) -> None:
         help="Server Address to be used to communicate with the clients",
         default="0.0.0.0:8080",
     )
-    parser.add_argument(
-        "--run_name",
-        action="store",
-        help="Name of the run, model checkpoints will be saved under a subfolder with this name",
-        required=True,
-    )
     args = parser.parse_args()
 
     config = load_config(args.config_path)
     log(INFO, f"Server Address: {args.server_address}")
-    main(config, args.server_address, args.run_name)
+    main(config, args.server_address)

research/flamby/fed_isic2019/fedavg/server.py

@@ -5,7 +5,6 @@
 from typing import Any, Dict
 
 import flwr as fl
-import torch
 from flamby.datasets.fed_isic2019 import Baseline
 from flwr.common.logger import log
 from flwr.server.client_manager import SimpleClientManager
@@ -22,7 +21,7 @@
 )
 
 
-def main(config: Dict[str, Any], server_address: str, checkpoint_stub: str, run_name: str, pretrain: bool) -> None:
+def main(config: Dict[str, Any], server_address: str, checkpoint_stub: str, run_name: str) -> None:
     # This function will be used to produce a config that is sent to each client to initialize their own environment
     fit_config_fn = partial(
         fit_config,
@@ -37,28 +36,6 @@ def main(config: Dict[str, Any], server_address: str, checkpoint_stub: str, run_
     client_manager = SimpleClientManager()
     model = Baseline()
 
-    log(INFO, f"if pretrain: {pretrain}")
-    if pretrain:
-        dir = (
-            "/ssd003/projects/aieng/public/FL_env/models/fed_isic2019/fedavg/hp_sweep_results/lr_0.001/"
-            + run_name
-            + "/server_best_model.pkl"
-        )
-        fedavg_model_state = torch.load(dir).state_dict()
-        model_state = model.state_dict()
-        matching_state = {}
-        for k, v in fedavg_model_state.items():
-            if k in model_state:
-                if v.size() == model_state[k].size():
-                    matching_state[k] = v
-                elif model_state[k].size()[1:] == v.size()[1:]:
-                    repeat = model_state[k].size()[0] // v.size()[0]
-                    original_size = tuple([1] * (len(model_state[k].size()) - 1))
-                    matching_state[k] = v.repeat((repeat,) + original_size)
-        log(INFO, f"matching state: {len(matching_state)}")
-        model_state.update(matching_state)
-        model.load_state_dict(model_state)
-
     # Server performs simple FedAveraging as its server-side optimization strategy
     strategy = FedAvg(
         min_fit_clients=config["n_clients"],
@@ -125,4 +102,4 @@ def main(config: Dict[str, Any], server_address: str, checkpoint_stub: str, run_
 
     config = load_config(args.config_path)
     log(INFO, f"Server Address: {args.server_address}")
-    main(config, args.server_address, args.artifact_dir, args.run_name, args.pretrain)
+    main(config, args.server_address, args.artifact_dir, args.run_name)

research/flamby/fed_isic2019/fenda/client.py

@@ -33,7 +33,7 @@ def __init__(
         loss_meter_type: LossMeterType = LossMeterType.AVERAGE,
         metric_meter_type: MetricMeterType = MetricMeterType.ACCUMULATION,
         checkpointer: Optional[TorchCheckpointer] = None,
-        type_run: str = "cos_sim",
+        type_run: str = "vanilla",
     ) -> None:
         super().__init__(
             data_path=data_path,

research/flamby/fed_isic2019/fenda/run_fold_experiment.slrm

@@ -139,7 +139,6 @@ do
             --client_number ${c} \
             --learning_rate ${CLIENT_LR} \
             --server_address ${SERVER_ADDRESS} \
-            --type_run cos_sim \
             > ${CLIENT_LOG_PATH} 2>&1 &
     done
 

research/flamby/fed_isic2019/fenda/server.py

@@ -20,7 +20,7 @@
 )
 
 
-def main(config: Dict[str, Any], server_address: str, run_name: str) -> None:
+def main(config: Dict[str, Any], server_address: str) -> None:
     # This function will be used to produce a config that is sent to each client to initialize their own environment
     fit_config_fn = partial(
         fit_config,
@@ -87,4 +87,4 @@ def main(config: Dict[str, Any], server_address: str, run_name: str) -> None:
 
     config = load_config(args.config_path)
     log(INFO, f"Server Address: {args.server_address}")
-    main(config, args.server_address, args.run_name)
+    main(config, args.server_address)

research/flamby/fed_ixi/fenda/client.py

@@ -33,7 +33,7 @@ def __init__(
         loss_meter_type: LossMeterType = LossMeterType.AVERAGE,
         metric_meter_type: MetricMeterType = MetricMeterType.ACCUMULATION,
         checkpointer: Optional[TorchCheckpointer] = None,
-        type_run: str = "cos_sim",
+        type_run: str = "vanilla",
     ) -> None:
 
         super().__init__(