diff --git a/scripts/train/train.py b/scripts/train/train.py
index c9e2d67bf4..3cf3d9551d 100644
--- a/scripts/train/train.py
+++ b/scripts/train/train.py
@@ -504,7 +504,7 @@ def main(cfg: DictConfig) -> Trainer:
         precision=train_cfg.precision,
         algorithms=algorithms,
         device_train_microbatch_size=train_cfg.device_train_microbatch_size,
-        fsdp_config=fsdp_config,
+        parallelism_config={'fsdp': fsdp_config},
         save_folder=train_cfg.save_folder,
         save_filename=save_filename,
         save_latest_filename=save_latest_filename,
diff --git a/setup.py b/setup.py
index 78182976d4..0556050de9 100644
--- a/setup.py
+++ b/setup.py
@@ -54,7 +54,7 @@
 ]
 
 install_requires = [
-    'mosaicml[libcloud,wandb,oci,gcs]>=0.22.0,<0.23',
+    'mosaicml[libcloud,wandb,oci,gcs]>=0.23.0,<0.24',
     'mlflow>=2.12.1,<2.13',
     'accelerate>=0.25,<0.26',  # for HF inference `device_map`
     'transformers>=4.40,<4.41',
@@ -92,14 +92,14 @@
 ]
 
 extra_deps['databricks'] = [
-    'mosaicml[databricks]>=0.22.0,<0.23',
+    'mosaicml[databricks]>=0.23.0,<0.24',
     'databricks-sql-connector>=3,<4',
     'databricks-connect==14.1.0',
     'lz4>=4,<5',
 ]
 
 extra_deps['tensorboard'] = [
-    'mosaicml[tensorboard]>=0.22.0,<0.23',
+    'mosaicml[tensorboard]>=0.23.0,<0.24',
 ]
 
 # Flash 2 group kept for backwards compatibility
@@ -110,7 +110,7 @@
 extra_deps['gpu'] = copy.deepcopy(extra_deps['gpu-flash2'])
 
 extra_deps['peft'] = [
-    'mosaicml[peft]>=0.22.0,<0.23',
+    'mosaicml[peft]>=0.23.0,<0.24',
 ]
 
 extra_deps['openai'] = [
diff --git a/tests/models/test_model.py b/tests/models/test_model.py
index a62a7dd114..2f93b1d3ce 100644
--- a/tests/models/test_model.py
+++ b/tests/models/test_model.py
@@ -13,10 +13,15 @@
 import torch.nn as nn
 from accelerate import init_empty_weights
 from composer.core.precision import Precision, get_precision_context
+from composer.distributed.dist_strategy import prepare_fsdp_module
 from composer.models.huggingface import maybe_get_underlying_model
 from composer.optim import DecoupledAdamW
-from composer.trainer.dist_strategy import prepare_fsdp_module
-from composer.utils import dist, get_device, reproducibility
+from composer.utils import (
+    FSDPConfig,
+    dist,
+    get_device,
+    reproducibility,
+)
 from omegaconf import DictConfig, ListConfig
 from omegaconf import OmegaConf as om
 from transformers import (
@@ -2538,7 +2543,14 @@ def test_hf_init(
         betas=(0.9, 0.99),
     )
 
-    prepare_fsdp_module(model, optimizer, fsdp_config, precision, device, False)
+    prepare_fsdp_module(
+        model,
+        optimizer,
+        FSDPConfig(**fsdp_config),
+        precision,
+        device,
+        False,
+    )
 
     model = HuggingFaceModelWithFSDP(model, tokenizer)