fix: fixing linting issues, updating gh action

.github/workflows/hgf-sync-main.yml

@@ -1,5 +1,7 @@
 # workflow that syncs the main branch to the Hugging Face Hub (Huggingface Spaces)
-#CREDIT Copied from Hugging Face, Inc.
+# only syncs if the build and lint is also fine
+# CREDIT: Adapted from Hugging Face, Inc.
+## see https://huggingface.co/docs/hub/spaces-github-actions
 
 name: HGF Hub Sync (Main)
 # runs on pushes to the main branch and manually triggered workflows
@@ -9,8 +11,35 @@ on:
 
   workflow_dispatch:
 
+permissions:
+  contents: read
+
 # jobs to run
 jobs:
+  
+  # build job that installs dependencies and lints
+  build:
+    runs-on: ubuntu-latest
+    steps:
+    # checkout the repository
+    - uses: actions/checkout@v3
+    # set up python 3.10
+    - name: Set up Python 3.10
+      uses: actions/setup-python@v3
+      with:
+        python-version: "3.10"
+    # install dependencies from requirements.txt
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install pylint black
+        if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+    # lint and format all files with black (as defined in pyproject.toml)
+    - name: Lint & Fix with Black
+      run: black .
+    # lint with pylint
+    - name: Lint with Pylint
+      run: pylint .
   # sync job
   sync-to-hub:
     runs-on: ubuntu-latest

backend/controller.py

@@ -40,15 +40,13 @@ def interference(
                 raise RuntimeError("There was an error in the selected XAI approach.")
 
         # call the explained chat function
-        prompt_output, history_output, xai_graphic, xai_markup = (
-            explained_chat(
-                model=godel,
-                xai=xai,
-                message=prompt,
-                history=history,
-                system_prompt=system_prompt,
-                knowledge=knowledge,
-            )
+        prompt_output, history_output, xai_graphic, xai_markup = explained_chat(
+            model=godel,
+            xai=xai,
+            message=prompt,
+            history=history,
+            system_prompt=system_prompt,
+            knowledge=knowledge,
         )
     # if no (or invalid) XAI approach is selected call the vanilla chat function
     else:

explanation/interpret_shap.py

@@ -1,8 +1,5 @@
 # interpret module that implements the interpretability method
 # external imports
-import seaborn as sns
-import matplotlib.pyplot as plt
-import numpy as np
 from shap import models, maskers, plots, PartitionExplainer
 import torch
 
@@ -62,4 +59,4 @@ def create_graphic(shap_values):
     graphic_html = plots.text(shap_values, display=False)
 
     # return the html graphic as string
-    return str(graphic_html)
+    return str(graphic_html)

explanation/visualize.py

@@ -1,9 +1,5 @@
 # visualization module that creates an attention visualization using BERTViz
 
-# external imports
-import matplotlib.pyplot as plt
-import seaborn as sns
-import numpy as np
 
 # internal imports
 from utils import formatting as fmt
@@ -41,4 +37,3 @@ def chat_explained(model, prompt):
     marked_text = markup_text(encoder_text, averaged_attention, variant="visualizer")
 
     return response_text, "", marked_text
-

main.py

@@ -180,10 +180,7 @@ with gr.Blocks(
                             " scripts: hieroglyphs, Demotic, and Greek."
                         ),
                     ],
-                    [
-                        "Does money buy happiness?",
-                        ""
-                    ],
+                    ["Does money buy happiness?", ""],
                 ],
                 inputs=[user_prompt, knowledge_input],
             )

utils/formatting.py

@@ -69,9 +69,11 @@ def format_tokens(tokens: list):
 def flatten_attribution(values: ndarray, axis: int = 0):
     return np.sum(values, axis=axis)
 
+
 def flatten_attention(values: ndarray, axis: int = 0):
     return np.mean(values, axis=axis)
 
+
 def avg_attention(attention_values):
     attention = attention_values.cross_attentions[0][0].detach().numpy()
     return np.mean(attention, axis=0)