diff --git a/lm_engine/finetune.py b/lm_engine/finetune.py
index 2139915d6..a266fd66b 100644
--- a/lm_engine/finetune.py
+++ b/lm_engine/finetune.py
@@ -6,7 +6,6 @@
 
 import torch
 from torch.distributed.tensor.parallel import loss_parallel
-from transformers import set_seed
 
 from .arguments import TrainingArgs, get_args
 from .checkpointing import ensure_last_checkpoint_is_saved, load_checkpoint_for_training, save_checkpoint
@@ -29,6 +28,7 @@
     StepTracker,
     TorchProfiler,
     init_distributed,
+    set_seed,
     setup_tf32,
 )
 
diff --git a/lm_engine/pretrain.py b/lm_engine/pretrain.py
index dbd44cbb3..23c5634a4 100644
--- a/lm_engine/pretrain.py
+++ b/lm_engine/pretrain.py
@@ -12,7 +12,6 @@
 from torch.distributed.pipelining.schedules import _PipelineSchedule
 from torch.distributed.tensor.parallel import loss_parallel
 from torch.utils.data import DataLoader
-from transformers import set_seed
 
 from .arguments import DistillationArgs, TrainingArgs, get_args
 from .checkpointing import ensure_last_checkpoint_is_saved, load_checkpoint_for_training, save_checkpoint
@@ -39,6 +38,7 @@
     is_torchao_available,
     log_environment,
     log_rank_0,
+    set_seed,
     setup_tf32,
 )
 
diff --git a/lm_engine/utils/__init__.py b/lm_engine/utils/__init__.py
index 9d8f7a504..99db5a27b 100644
--- a/lm_engine/utils/__init__.py
+++ b/lm_engine/utils/__init__.py
@@ -34,6 +34,7 @@
 from .parallel import ProcessGroupManager, get_pipeline_stage_ids_on_current_rank, run_rank_n
 from .profiler import TorchProfiler
 from .pydantic import BaseArgs
+from .random import set_seed
 from .safetensors import SafeTensorsWeightsManager
 from .step_tracker import StepTracker
 from .tracking import ExperimentsTracker, ProgressBar
diff --git a/lm_engine/utils/random.py b/lm_engine/utils/random.py
new file mode 100644
index 000000000..c9c0ae60e
--- /dev/null
+++ b/lm_engine/utils/random.py
@@ -0,0 +1,15 @@
+# **************************************************
+# Copyright (c) 2025, Mayank Mishra
+# **************************************************
+
+import random
+
+import numpy as np
+import torch
+
+
+def set_seed(seed: int) -> None:
+    random.seed(seed)
+    np.random.seed(seed)
+    torch.manual_seed(seed)
+    torch.cuda.manual_seed_all(seed)
diff --git a/tests/hf_models/multi_gpu/tensor_parallel/tensor_parallel_forward.py b/tests/hf_models/multi_gpu/tensor_parallel/tensor_parallel_forward.py
index fce40795e..431ebba7d 100644
--- a/tests/hf_models/multi_gpu/tensor_parallel/tensor_parallel_forward.py
+++ b/tests/hf_models/multi_gpu/tensor_parallel/tensor_parallel_forward.py
@@ -7,12 +7,17 @@
 
 import torch
 import torch.distributed
-from transformers import set_seed
 
 from lm_engine.enums import Kernel
 from lm_engine.hf_models import GPTBaseConfig, get_model_parallel_class
 from lm_engine.kernels import enable_kernels
-from lm_engine.utils import Communication, ProcessGroupManager, SafeTensorsWeightsManager, string_to_torch_dtype
+from lm_engine.utils import (
+    Communication,
+    ProcessGroupManager,
+    SafeTensorsWeightsManager,
+    set_seed,
+    string_to_torch_dtype,
+)
 
 from ...test_common import TestCommons
 
diff --git a/tests/hf_models/single_gpu/gpt_base_test.py b/tests/hf_models/single_gpu/gpt_base_test.py
index 579437443..9477c212c 100644
--- a/tests/hf_models/single_gpu/gpt_base_test.py
+++ b/tests/hf_models/single_gpu/gpt_base_test.py
@@ -6,10 +6,10 @@
 
 import torch
 from parameterized import parameterized
-from transformers import set_seed
 
 from lm_engine.enums import Kernel
 from lm_engine.kernels import enable_kernels
+from lm_engine.utils import set_seed
 
 from ..test_common import TestCommons
 
diff --git a/tests/hf_models/single_gpu/multihead_latent_attention_test.py b/tests/hf_models/single_gpu/multihead_latent_attention_test.py
index 4accd572b..f20ad9281 100644
--- a/tests/hf_models/single_gpu/multihead_latent_attention_test.py
+++ b/tests/hf_models/single_gpu/multihead_latent_attention_test.py
@@ -4,11 +4,11 @@
 
 import torch
 from parameterized import parameterized
-from transformers import set_seed
 
 from lm_engine.enums import Kernel
 from lm_engine.hf_models import GPTBaseConfig
 from lm_engine.kernels import enable_kernels
+from lm_engine.utils import set_seed
 
 from ..test_common import TestCommons
 
diff --git a/tests/hf_models/single_gpu/scattermoe_test.py b/tests/hf_models/single_gpu/scattermoe_test.py
index 0bfedaf79..b0234966e 100644
--- a/tests/hf_models/single_gpu/scattermoe_test.py
+++ b/tests/hf_models/single_gpu/scattermoe_test.py
@@ -4,10 +4,10 @@
 
 import torch
 from parameterized import parameterized
-from transformers import set_seed
 
 from lm_engine.enums import Kernel
 from lm_engine.kernels import enable_kernels
+from lm_engine.utils import set_seed
 
 from ..test_common import TestCommons
 
diff --git a/tests/training/params_group/efficient_init_test.py b/tests/training/params_group/efficient_init_test.py
index 183205e6a..31dfac81c 100644
--- a/tests/training/params_group/efficient_init_test.py
+++ b/tests/training/params_group/efficient_init_test.py
@@ -5,13 +5,12 @@
 import os
 
 import torch
-from transformers import set_seed
 
 from lm_engine.arguments import UnshardingArgs
 from lm_engine.checkpointing import load_checkpoint_and_unshard, save_checkpoint
 from lm_engine.distributed import wrap_model_container_for_distributed_training
 from lm_engine.model_wrapper import get_model_container
-from lm_engine.utils import ProcessGroupManager, load_yaml
+from lm_engine.utils import ProcessGroupManager, load_yaml, set_seed
 
 from ..test_commons import TestCommons