diff --git a/examples/benchmarks/bert/src/bert_layers.py b/examples/benchmarks/bert/src/bert_layers.py
index b2c5ff267..085c16c0e 100644
--- a/examples/benchmarks/bert/src/bert_layers.py
+++ b/examples/benchmarks/bert/src/bert_layers.py
@@ -63,8 +63,8 @@
 IMPL_USE_FLASH2 = False
 # Import Flash Attention 2, which supports ALiBi https://github.com/Dao-AILab/flash-attention
 try:
-    from flash_attn import flash_attn_qkvpacked_func # type: ignore
-    installed_version = importlib.metadata.version('flash_attn') # type: ignore
+    from flash_attn import flash_attn_qkvpacked_func  # type: ignore
+    installed_version = importlib.metadata.version('flash_attn')  # type: ignore
     if installed_version < '2.4.2':
         raise ImportError('newer version of flash_attn required (>= 2.4.2)')
     IMPL_USE_FLASH2 = True
@@ -278,7 +278,7 @@ def forward(self, hidden_states: torch.Tensor, cu_seqlens: torch.Tensor,
 
                     attention = flash_attn_qkvpacked_func(
                         qkv, dropout_p=self.p_dropout, alibi_slopes=slopes)
-                    attention = attention.to(orig_dtype) # type: ignore
+                    attention = attention.to(orig_dtype)  # type: ignore
                     bias = bias.to(bias_dtype)
                 else:
                     attention = flash_attn_qkvpacked_func(
@@ -302,14 +302,14 @@ def forward(self, hidden_states: torch.Tensor, cu_seqlens: torch.Tensor,
                     bias_dtype = bias.dtype
                     bias = bias.to(half)
                     attention = flash_attn_qkvpacked_func(qkv, bias)
-                    attention = attention.to(orig_dtype) # type: ignore
+                    attention = attention.to(orig_dtype)  # type: ignore
                     bias = bias.to(bias_dtype)
                 else:
                     attention = flash_attn_qkvpacked_func(qkv, bias)
 
         # attn_mask is 1 for attend and 0 for don't attend
         attention = bert_padding_module.unpad_input_only(
-            attention, # type: ignore
+            attention,  # type: ignore
             torch.squeeze(attn_mask) == 1)
         return rearrange(attention, 'nnz h d -> nnz (h d)')
 
diff --git a/examples/end-to-end-examples/support_chatbot/__init__.py b/examples/end-to-end-examples/support_chatbot/__init__.py
index 642d7387a..cb0d03b22 100644
--- a/examples/end-to-end-examples/support_chatbot/__init__.py
+++ b/examples/end-to-end-examples/support_chatbot/__init__.py
@@ -1,3 +1,2 @@
 # Copyright 2022 MosaicML Examples authors
 # SPDX-License-Identifier: Apache-2.0
-
diff --git a/examples/end-to-end-examples/support_chatbot/mcli_yamls/conversion/convert_txt_to_stream.yaml b/examples/end-to-end-examples/support_chatbot/mcli_yamls/conversion/convert_txt_to_stream.yaml
index 506952be1..f2d63083e 100644
--- a/examples/end-to-end-examples/support_chatbot/mcli_yamls/conversion/convert_txt_to_stream.yaml
+++ b/examples/end-to-end-examples/support_chatbot/mcli_yamls/conversion/convert_txt_to_stream.yaml
@@ -11,7 +11,7 @@ integrations:
 # Clone the examples repository so that we have access to the code in sec_10k_qa
 - integration_type: git_repo
   git_repo: YOUR_GITHUB_USERNAME/examples
-  #git_branch: support-bot
+  # git_branch: support-bot
   ssh_clone: false # Should be true if using a private repo
   path: /workspace/examples # Tell MCLI what path to clone the repo to
 
diff --git a/examples/end-to-end-examples/support_chatbot/mcli_yamls/finetune/finetune_composer_codebase.yaml b/examples/end-to-end-examples/support_chatbot/mcli_yamls/finetune/finetune_composer_codebase.yaml
index 526dcc9cc..6644a8aaa 100644
--- a/examples/end-to-end-examples/support_chatbot/mcli_yamls/finetune/finetune_composer_codebase.yaml
+++ b/examples/end-to-end-examples/support_chatbot/mcli_yamls/finetune/finetune_composer_codebase.yaml
@@ -5,7 +5,7 @@ compute:
 
   ## These configurations are optional
   # cluster: r0z0 # Name of the cluster to use for this run
-  #gpu_type: h100_80gb # Type of GPU to use.
+  # gpu_type: h100_80gb # Type of GPU to use.
 
 integrations:
 # Clone and install the llm-foundry repo so we can run scripts from it
@@ -25,7 +25,7 @@ command: |
   cd llm-foundry/scripts
   composer train/train.py /mnt/config/parameters.yaml || (echo "Command failed - killing python" && pkill python && exit 1)
 
-#image: "mosaicml/llm-foundry:2.0.1_cu118-latest"
+# image: "mosaicml/llm-foundry:2.0.1_cu118-latest"
 image: mosaicml/pytorch:1.13.1_cu117-python3.10-ubuntu20.04
 
 # The parameters section is mounted to /mnt/config/parameters.yaml in the container
@@ -60,7 +60,7 @@ parameters:
   model:
     name: hf_causal_lm
     init_device: mixed # Initially only create the model on CPU once per node to reduce system memory requirements
-    pretrained_model_name_or_path: mosaicml/mpt-30b-chat # This can be changed to other models from the HuggingFace model hub
+    pretrained_model_name_or_path: mosaicml/mpt-30b-chat # This can be changed to other HF models
     pretrained: true # If false, will just load the model architecture and randomly initialize the weights
     config_overrides: # Override the default model config (comment this out if you change the model from MPT)
       attn_config:
diff --git a/examples/end-to-end-examples/support_chatbot/repo_downloader.py b/examples/end-to-end-examples/support_chatbot/repo_downloader.py
index 85d9cd686..889781235 100644
--- a/examples/end-to-end-examples/support_chatbot/repo_downloader.py
+++ b/examples/end-to-end-examples/support_chatbot/repo_downloader.py
@@ -91,7 +91,9 @@ def prepare_output_file(self, file_path: str) -> str:
         return output_file
 
     def file_to_txt(self, file_path: (str)) -> None:
-        """Given the file_path of a file in cloned repository, downloads it to a
+        """Given the file_path of a file in cloned repository, downloads it to
+        a.
+
         .txt file and saves it in the same directory structure in.
 
         /scripts/train/support_chatbot/retrieval_data/{self.repo_name}