cciwon-code-review-cli 2.0.1 → 2.0.3

package/bin/code-review.js

@@ -16,7 +16,7 @@ program
   .version('1.0.0');
 
 // 서버 URL 설정 (환경변수 또는 기본값)
-const SERVER_URL = process.env.CODE_REVIEW_SERVER || 'http://192.168.10.152:8000';
+const SERVER_URL = process.env.CODE_REVIEW_SERVER || 'http://211.56.247.71:8000';
 const client = new CodeReviewClient(SERVER_URL);
 
 // ==========================================

package/lib/chat-mode.js

@@ -145,16 +145,21 @@ class ChatMode {
    */
   async reviewFile(filePath) {
     try {
+      console.log(chalk.bold(`\n=== ${filePath} ===\n`));
+
       // 파일 내용 읽기
+      console.log(chalk.gray(`📖 파일 읽는 중...`));
       const content = await this.fileReader.readFile(filePath);
+      console.log(chalk.gray(`✓ 파일 읽기 완료 (${content.length}자)`));
 
       // 언어 감지
       const language = this.detectLanguage(filePath);
-
-      console.log(chalk.bold(`\n=== ${filePath} ===\n`));
+      console.log(chalk.gray(`🔍 언어: ${language}`));
 
       // 서버에 리뷰 요청
+      console.log(chalk.gray(`🚀 서버로 리뷰 요청 중...`));
       const result = await this.client.reviewCode(content, language);
+      console.log(chalk.gray(`✓ 리뷰 완료\n`));
 
       // 결과 출력
       console.log(result.review);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cciwon-code-review-cli",
-  "version": "2.0.1",
+  "version": "2.0.3",
   "description": "AI-powered code review CLI tool using Qwen3-Coder-30B model with IP whitelist support",
   "main": "lib/api-client.js",
   "bin": {

package/unsloth_compiled_cache/AqlmLoraLinear_peft_forward.py

@@ -0,0 +1,88 @@
+"""
+2025.12.6
+2025.12.7
+4.57.1
+0.24.0
+__UNSLOTH_VERSIONING__
+"""
+
+# Unsloth auto generated code
+# Copyright 2023-present Daniel Han-Chen, Michael Han-Chen & the Unsloth team. All rights reserved.
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+
+torch_compile_options = {'epilogue_fusion': True, 'max_autotune': False, 'shape_padding': True, 'trace.enabled': False, 'triton.cudagraphs': False, 'debug': False, 'dce': True, 'memory_planning': True, 'coordinate_descent_tuning': False, 'trace.graph_diagram': False, 'compile_threads': 32, 'group_fusion': True, 'disable_progress': True, 'verbose_progress': False, 'triton.multi_kernel': 0, 'triton.use_block_ptr': False, 'triton.enable_persistent_tma_matmul': True, 'triton.autotune_at_compile_time': False, 'triton.cooperative_reductions': False, 'cuda.compile_opt_level': '-O2', 'cuda.enable_cuda_lto': True, 'combo_kernels': False, 'benchmark_combo_kernel': True, 'combo_kernel_foreach_dynamic_shapes': True}
+from torch import Tensor
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+from peft.tuners.lora.aqlm import (torch)
+
+
+torch_addmm = torch.addmm
+torch_add   = torch.add
+# @torch.compile(fullgraph = False, dynamic = True, options = torch_compile_options)
+def lora_forward(result, lora_A, lora_B, dropout, x, scaling):
+    # Use result.dtype (bfloat16 from base layer) since x may have been cast to float32
+    # by _cast_input_dtype when autocast is disabled
+    target_dtype = result.dtype
+    xA = dropout(x).to(target_dtype) @ lora_A.weight.to(target_dtype).t()
+    # output = result + scaling * xA @ lora_B.weight.t()
+    shape = result.shape
+    output = torch_addmm(
+        result.view(-1, shape[-1]),
+        xA.view(-1, xA.shape[-1]),
+        lora_B.weight.to(target_dtype).t(),
+        alpha = scaling,
+        beta = 1,
+    ).view(shape)
+
+    bias = lora_B.bias
+    if bias is not None:
+        output = torch_add(
+            output,
+            bias.to(target_dtype),
+            alpha = scaling,
+        )
+    return output
+pass
+
+def unsloth_forward(self, x: torch.Tensor):
+    # note: logic differs from default Linear because merging is not supported
+    result = self.base_layer(x)
+
+    if self.disable_adapters:
+        return result
+
+    for active_adapter in self.active_adapters:
+        if active_adapter not in self.lora_A.keys():
+            continue
+        lora_A = self.lora_A[active_adapter]
+        lora_B = self.lora_B[active_adapter]
+        dropout = self.lora_dropout[active_adapter]
+        scaling = self.scaling[active_adapter]
+
+        requires_conversion = not torch.is_autocast_enabled()
+        if requires_conversion:
+            expected_dtype = result.dtype
+            x = self._cast_input_dtype(x, lora_A.weight.dtype)
+
+        output = lora_B(lora_A(dropout(x)))
+        if requires_conversion:
+            output = output.to(expected_dtype)
+        output = output * scaling
+        result += output
+    return result

package/unsloth_compiled_cache/AwqLoraLinear_peft_forward.py

@@ -0,0 +1,87 @@
+"""
+2025.12.6
+2025.12.7
+4.57.1
+0.24.0
+__UNSLOTH_VERSIONING__
+"""
+
+# Unsloth auto generated code
+# Copyright 2023-present Daniel Han-Chen, Michael Han-Chen & the Unsloth team. All rights reserved.
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+
+torch_compile_options = {'epilogue_fusion': True, 'max_autotune': False, 'shape_padding': True, 'trace.enabled': False, 'triton.cudagraphs': False, 'debug': False, 'dce': True, 'memory_planning': True, 'coordinate_descent_tuning': False, 'trace.graph_diagram': False, 'compile_threads': 32, 'group_fusion': True, 'disable_progress': True, 'verbose_progress': False, 'triton.multi_kernel': 0, 'triton.use_block_ptr': False, 'triton.enable_persistent_tma_matmul': True, 'triton.autotune_at_compile_time': False, 'triton.cooperative_reductions': False, 'cuda.compile_opt_level': '-O2', 'cuda.enable_cuda_lto': True, 'combo_kernels': False, 'benchmark_combo_kernel': True, 'combo_kernel_foreach_dynamic_shapes': True}
+from torch import Tensor
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+from peft.tuners.lora.awq import (torch)
+
+
+torch_addmm = torch.addmm
+torch_add   = torch.add
+# @torch.compile(fullgraph = False, dynamic = True, options = torch_compile_options)
+def lora_forward(result, lora_A, lora_B, dropout, x, scaling):
+    # Use result.dtype (bfloat16 from base layer) since x may have been cast to float32
+    # by _cast_input_dtype when autocast is disabled
+    target_dtype = result.dtype
+    xA = dropout(x).to(target_dtype) @ lora_A.weight.to(target_dtype).t()
+    # output = result + scaling * xA @ lora_B.weight.t()
+    shape = result.shape
+    output = torch_addmm(
+        result.view(-1, shape[-1]),
+        xA.view(-1, xA.shape[-1]),
+        lora_B.weight.to(target_dtype).t(),
+        alpha = scaling,
+        beta = 1,
+    ).view(shape)
+
+    bias = lora_B.bias
+    if bias is not None:
+        output = torch_add(
+            output,
+            bias.to(target_dtype),
+            alpha = scaling,
+        )
+    return output
+pass
+
+def unsloth_forward(self, x: torch.Tensor):
+    result = self.quant_linear_module(x)
+
+    if self.disable_adapters:
+        return result
+
+    for active_adapter in self.active_adapters:
+        if active_adapter not in self.lora_A.keys():
+            continue
+        lora_A = self.lora_A[active_adapter]
+        lora_B = self.lora_B[active_adapter]
+        dropout = self.lora_dropout[active_adapter]
+        scaling = self.scaling[active_adapter]
+
+        requires_conversion = not torch.is_autocast_enabled()
+        if requires_conversion:
+            expected_dtype = result.dtype
+            x = self._cast_input_dtype(x, lora_A.weight.dtype)
+
+        output = lora_B(lora_A(dropout(x)))
+        if requires_conversion:
+            output = output.to(expected_dtype)
+        output = output * scaling
+        result = result + output
+    return result

package/unsloth_compiled_cache/BatchNorm1d.py

@@ -0,0 +1,117 @@
+"""
+2025.12.6
+2025.12.7
+4.57.1
+0.24.0
+__UNSLOTH_VERSIONING__
+"""
+
+# Unsloth auto generated code
+# Copyright 2023-present Daniel Han-Chen, Michael Han-Chen & the Unsloth team. All rights reserved.
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+
+import os
+import torch
+import importlib.util
+import math
+if importlib.util.find_spec("unsloth_studio") is None:
+    UNSLOTH_STUDIO_ENABLED = False
+else:
+    UNSLOTH_STUDIO_ENABLED = os.environ.get("UNSLOTH_STUDIO_DISABLED", "0") == "0"
+pass
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+import math
+
+UNSLOTH_ENABLE_LOGGING = os.environ.get("UNSLOTH_ENABLE_LOGGING", "0") == "1"
+UNSLOTH_ENABLE_CCE = os.environ.get("UNSLOTH_ENABLE_CCE", "1") == "1"
+UNSLOTH_COMPILE_DISABLE = os.environ.get("UNSLOTH_COMPILE_DISABLE", "0") in ("1", "partial",)
+
+import logging
+logger_compiler = logging.getLogger(__name__)
+if UNSLOTH_ENABLE_LOGGING:
+    logger_compiler.setLevel(logging.DEBUG)
+
+global INFERENCE_RUNS
+INFERENCE_RUNS = 0
+
+try:
+    import torch._dynamo.eval_frame as torch_dynamo_eval_frame
+    torch_dynamo_eval_frame._stance.stance
+    torch_compiler_set_stance = torch.compiler.set_stance
+except:
+    torch_dynamo_eval_frame = None
+    torch_compiler_set_stance = None
+pass
+
+from unsloth_zoo import DEVICE_TYPE_TORCH, DEVICE_COUNT
+
+torch_compile_options = {'epilogue_fusion': True, 'max_autotune': False, 'shape_padding': True, 'trace.enabled': False, 'triton.cudagraphs': False, 'debug': False, 'dce': True, 'memory_planning': True, 'coordinate_descent_tuning': False, 'trace.graph_diagram': False, 'compile_threads': 32, 'group_fusion': True, 'disable_progress': True, 'verbose_progress': False, 'triton.multi_kernel': 0, 'triton.use_block_ptr': False, 'triton.enable_persistent_tma_matmul': True, 'triton.autotune_at_compile_time': False, 'triton.cooperative_reductions': False, 'cuda.compile_opt_level': '-O2', 'cuda.enable_cuda_lto': True, 'combo_kernels': False, 'benchmark_combo_kernel': True, 'combo_kernel_foreach_dynamic_shapes': True}
+from torch import Tensor
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+from transformers.models.qwen3_moe.modeling_qwen3_moe import (F, nn)
+
+def forward(self, input: Tensor) -> Tensor:
+    self._check_input_dim(input)
+
+    # exponential_average_factor is set to self.momentum
+    # (when it is available) only so that it gets updated
+    # in ONNX graph when this node is exported to ONNX.
+    if self.momentum is None:
+        exponential_average_factor = 0.0
+    else:
+        exponential_average_factor = self.momentum
+
+    if self.training and self.track_running_stats:
+        # TODO: if statement only here to tell the jit to skip emitting this when it is None
+        if self.num_batches_tracked is not None:  # type: ignore[has-type]
+            self.num_batches_tracked.add_(1)  # type: ignore[has-type]
+            if self.momentum is None:  # use cumulative moving average
+                exponential_average_factor = 1.0 / float(self.num_batches_tracked)
+            else:  # use exponential moving average
+                exponential_average_factor = self.momentum
+
+    r"""
+    Decide whether the mini-batch stats should be used for normalization rather than the buffers.
+    Mini-batch stats are used in training mode, and in eval mode when buffers are None.
+    """
+    if self.training:
+        bn_training = True
+    else:
+        bn_training = (self.running_mean is None) and (self.running_var is None)
+
+    r"""
+    Buffers are only updated if they are to be tracked and we are in training mode. Thus they only need to be
+    passed when the update should occur (i.e. in training mode when they are tracked), or when buffer stats are
+    used for normalization (i.e. in eval mode when buffers are not None).
+    """
+    return F.batch_norm(
+        input,
+        # If buffers are not to be tracked, ensure that they won't be updated
+        (
+            self.running_mean
+            if not self.training or self.track_running_stats
+            else None
+        ),
+        self.running_var if not self.training or self.track_running_stats else None,
+        self.weight,
+        self.bias,
+        bn_training,
+        exponential_average_factor,
+        self.eps,
+    ).to(input.dtype).to(input.dtype)

package/unsloth_compiled_cache/BatchNorm2d.py

@@ -0,0 +1,117 @@
+"""
+2025.12.6
+2025.12.7
+4.57.1
+0.24.0
+__UNSLOTH_VERSIONING__
+"""
+
+# Unsloth auto generated code
+# Copyright 2023-present Daniel Han-Chen, Michael Han-Chen & the Unsloth team. All rights reserved.
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+
+import os
+import torch
+import importlib.util
+import math
+if importlib.util.find_spec("unsloth_studio") is None:
+    UNSLOTH_STUDIO_ENABLED = False
+else:
+    UNSLOTH_STUDIO_ENABLED = os.environ.get("UNSLOTH_STUDIO_DISABLED", "0") == "0"
+pass
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+import math
+
+UNSLOTH_ENABLE_LOGGING = os.environ.get("UNSLOTH_ENABLE_LOGGING", "0") == "1"
+UNSLOTH_ENABLE_CCE = os.environ.get("UNSLOTH_ENABLE_CCE", "1") == "1"
+UNSLOTH_COMPILE_DISABLE = os.environ.get("UNSLOTH_COMPILE_DISABLE", "0") in ("1", "partial",)
+
+import logging
+logger_compiler = logging.getLogger(__name__)
+if UNSLOTH_ENABLE_LOGGING:
+    logger_compiler.setLevel(logging.DEBUG)
+
+global INFERENCE_RUNS
+INFERENCE_RUNS = 0
+
+try:
+    import torch._dynamo.eval_frame as torch_dynamo_eval_frame
+    torch_dynamo_eval_frame._stance.stance
+    torch_compiler_set_stance = torch.compiler.set_stance
+except:
+    torch_dynamo_eval_frame = None
+    torch_compiler_set_stance = None
+pass
+
+from unsloth_zoo import DEVICE_TYPE_TORCH, DEVICE_COUNT
+
+torch_compile_options = {'epilogue_fusion': True, 'max_autotune': False, 'shape_padding': True, 'trace.enabled': False, 'triton.cudagraphs': False, 'debug': False, 'dce': True, 'memory_planning': True, 'coordinate_descent_tuning': False, 'trace.graph_diagram': False, 'compile_threads': 32, 'group_fusion': True, 'disable_progress': True, 'verbose_progress': False, 'triton.multi_kernel': 0, 'triton.use_block_ptr': False, 'triton.enable_persistent_tma_matmul': True, 'triton.autotune_at_compile_time': False, 'triton.cooperative_reductions': False, 'cuda.compile_opt_level': '-O2', 'cuda.enable_cuda_lto': True, 'combo_kernels': False, 'benchmark_combo_kernel': True, 'combo_kernel_foreach_dynamic_shapes': True}
+from torch import Tensor
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+from transformers.models.qwen3_moe.modeling_qwen3_moe import (F, nn)
+
+def forward(self, input: Tensor) -> Tensor:
+    self._check_input_dim(input)
+
+    # exponential_average_factor is set to self.momentum
+    # (when it is available) only so that it gets updated
+    # in ONNX graph when this node is exported to ONNX.
+    if self.momentum is None:
+        exponential_average_factor = 0.0
+    else:
+        exponential_average_factor = self.momentum
+
+    if self.training and self.track_running_stats:
+        # TODO: if statement only here to tell the jit to skip emitting this when it is None
+        if self.num_batches_tracked is not None:  # type: ignore[has-type]
+            self.num_batches_tracked.add_(1)  # type: ignore[has-type]
+            if self.momentum is None:  # use cumulative moving average
+                exponential_average_factor = 1.0 / float(self.num_batches_tracked)
+            else:  # use exponential moving average
+                exponential_average_factor = self.momentum
+
+    r"""
+    Decide whether the mini-batch stats should be used for normalization rather than the buffers.
+    Mini-batch stats are used in training mode, and in eval mode when buffers are None.
+    """
+    if self.training:
+        bn_training = True
+    else:
+        bn_training = (self.running_mean is None) and (self.running_var is None)
+
+    r"""
+    Buffers are only updated if they are to be tracked and we are in training mode. Thus they only need to be
+    passed when the update should occur (i.e. in training mode when they are tracked), or when buffer stats are
+    used for normalization (i.e. in eval mode when buffers are not None).
+    """
+    return F.batch_norm(
+        input,
+        # If buffers are not to be tracked, ensure that they won't be updated
+        (
+            self.running_mean
+            if not self.training or self.track_running_stats
+            else None
+        ),
+        self.running_var if not self.training or self.track_running_stats else None,
+        self.weight,
+        self.bias,
+        bn_training,
+        exponential_average_factor,
+        self.eps,
+    ).to(input.dtype).to(input.dtype)

package/unsloth_compiled_cache/BatchNorm3d.py

@@ -0,0 +1,117 @@
+"""
+2025.12.6
+2025.12.7
+4.57.1
+0.24.0
+__UNSLOTH_VERSIONING__
+"""
+
+# Unsloth auto generated code
+# Copyright 2023-present Daniel Han-Chen, Michael Han-Chen & the Unsloth team. All rights reserved.
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU Lesser General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+
+import os
+import torch
+import importlib.util
+import math
+if importlib.util.find_spec("unsloth_studio") is None:
+    UNSLOTH_STUDIO_ENABLED = False
+else:
+    UNSLOTH_STUDIO_ENABLED = os.environ.get("UNSLOTH_STUDIO_DISABLED", "0") == "0"
+pass
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+import math
+
+UNSLOTH_ENABLE_LOGGING = os.environ.get("UNSLOTH_ENABLE_LOGGING", "0") == "1"
+UNSLOTH_ENABLE_CCE = os.environ.get("UNSLOTH_ENABLE_CCE", "1") == "1"
+UNSLOTH_COMPILE_DISABLE = os.environ.get("UNSLOTH_COMPILE_DISABLE", "0") in ("1", "partial",)
+
+import logging
+logger_compiler = logging.getLogger(__name__)
+if UNSLOTH_ENABLE_LOGGING:
+    logger_compiler.setLevel(logging.DEBUG)
+
+global INFERENCE_RUNS
+INFERENCE_RUNS = 0
+
+try:
+    import torch._dynamo.eval_frame as torch_dynamo_eval_frame
+    torch_dynamo_eval_frame._stance.stance
+    torch_compiler_set_stance = torch.compiler.set_stance
+except:
+    torch_dynamo_eval_frame = None
+    torch_compiler_set_stance = None
+pass
+
+from unsloth_zoo import DEVICE_TYPE_TORCH, DEVICE_COUNT
+
+torch_compile_options = {'epilogue_fusion': True, 'max_autotune': False, 'shape_padding': True, 'trace.enabled': False, 'triton.cudagraphs': False, 'debug': False, 'dce': True, 'memory_planning': True, 'coordinate_descent_tuning': False, 'trace.graph_diagram': False, 'compile_threads': 32, 'group_fusion': True, 'disable_progress': True, 'verbose_progress': False, 'triton.multi_kernel': 0, 'triton.use_block_ptr': False, 'triton.enable_persistent_tma_matmul': True, 'triton.autotune_at_compile_time': False, 'triton.cooperative_reductions': False, 'cuda.compile_opt_level': '-O2', 'cuda.enable_cuda_lto': True, 'combo_kernels': False, 'benchmark_combo_kernel': True, 'combo_kernel_foreach_dynamic_shapes': True}
+from torch import Tensor
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+from typing import Any, List, Optional, Tuple, Union, Dict, Set, Callable
+from transformers.models.qwen3_moe.modeling_qwen3_moe import (F, nn)
+
+def forward(self, input: Tensor) -> Tensor:
+    self._check_input_dim(input)
+
+    # exponential_average_factor is set to self.momentum
+    # (when it is available) only so that it gets updated
+    # in ONNX graph when this node is exported to ONNX.
+    if self.momentum is None:
+        exponential_average_factor = 0.0
+    else:
+        exponential_average_factor = self.momentum
+
+    if self.training and self.track_running_stats:
+        # TODO: if statement only here to tell the jit to skip emitting this when it is None
+        if self.num_batches_tracked is not None:  # type: ignore[has-type]
+            self.num_batches_tracked.add_(1)  # type: ignore[has-type]
+            if self.momentum is None:  # use cumulative moving average
+                exponential_average_factor = 1.0 / float(self.num_batches_tracked)
+            else:  # use exponential moving average
+                exponential_average_factor = self.momentum
+
+    r"""
+    Decide whether the mini-batch stats should be used for normalization rather than the buffers.
+    Mini-batch stats are used in training mode, and in eval mode when buffers are None.
+    """
+    if self.training:
+        bn_training = True
+    else:
+        bn_training = (self.running_mean is None) and (self.running_var is None)
+
+    r"""
+    Buffers are only updated if they are to be tracked and we are in training mode. Thus they only need to be
+    passed when the update should occur (i.e. in training mode when they are tracked), or when buffer stats are
+    used for normalization (i.e. in eval mode when buffers are not None).
+    """
+    return F.batch_norm(
+        input,
+        # If buffers are not to be tracked, ensure that they won't be updated
+        (
+            self.running_mean
+            if not self.training or self.track_running_stats
+            else None
+        ),
+        self.running_var if not self.training or self.track_running_stats else None,
+        self.weight,
+        self.bias,
+        bn_training,
+        exponential_average_factor,
+        self.eps,
+    ).to(input.dtype).to(input.dtype)