Download my_pages/ica.py from verifiability-tutorial/verifiability: direct link, hf CLI and curl.
- Browser
- Download file 5.87 kB
-
https://huggingface.co/spaces/verifiability-tutorial/verifiability/resolve/main/my_pages/ica.py
- Command line
-
hf download hf://spaces/verifiability-tutorial/verifiability/my_pages/ica.py
-
curl -L -o ica.py https://huggingface.co/spaces/verifiability-tutorial/verifiability/resolve/main/my_pages/ica.py
5.87 kB
| import streamlit as st | |
| import matplotlib.pyplot as plt | |
| import numpy as np | |
| from utils import add_navigation, add_instruction_text, add_red_text | |
| plt.style.use('dark_background') | |
| def render(): | |
| add_navigation("txt_ica", "txt_multiverse") | |
| add_instruction_text( | |
| """ | |
| Explore the intention-convention-arbitrariness (ICA) framework.<br> | |
| Use the sliders to adjust the three dimensions and uncover various examples. | |
| """ | |
| ) | |
| # Initialize weights | |
| if "weights" not in st.session_state: | |
| st.session_state.weights = { | |
| "Intentional": 0.33, | |
| "Conventional": 0.33, | |
| "Arbitrary": 0.34 | |
| } | |
| # Keep track of previous weights | |
| if "prev_weights" not in st.session_state: | |
| st.session_state.prev_weights = st.session_state.weights.copy() | |
| w = st.session_state.weights | |
| prev_w = st.session_state.prev_weights | |
| # --- Three sliders --- | |
| col1, col2, col3 = st.columns(3) | |
| with col1: | |
| i_new = st.slider("Intentional", 0.0, 1.0, w["Intentional"], 0.01) | |
| with col2: | |
| c_new = st.slider("Conventional", 0.0, 1.0, w["Conventional"], 0.01) | |
| with col3: | |
| a_new = st.slider("Arbitrary", 0.0, 1.0, w["Arbitrary"], 0.01) | |
| # --- Adjust other sliders proportionally --- | |
| # Detect which slider changed | |
| if i_new != prev_w["Intentional"]: | |
| diff = i_new - prev_w["Intentional"] | |
| total_other = w["Conventional"] + w["Arbitrary"] | |
| if total_other > 0: | |
| w["Conventional"] -= diff * (w["Conventional"] / total_other) | |
| w["Arbitrary"] -= diff * (w["Arbitrary"] / total_other) | |
| w["Intentional"] = i_new | |
| st.session_state.prev_weights = w.copy() | |
| st.rerun() | |
| elif c_new != prev_w["Conventional"]: | |
| diff = c_new - prev_w["Conventional"] | |
| total_other = w["Intentional"] + w["Arbitrary"] | |
| if total_other > 0: | |
| w["Intentional"] -= diff * (w["Intentional"] / total_other) | |
| w["Arbitrary"] -= diff * (w["Arbitrary"] / total_other) | |
| w["Conventional"] = c_new | |
| st.session_state.prev_weights = w.copy() | |
| st.rerun() | |
| elif a_new != prev_w["Arbitrary"]: | |
| diff = a_new - prev_w["Arbitrary"] | |
| total_other = w["Intentional"] + w["Conventional"] | |
| if total_other > 0: | |
| w["Intentional"] -= diff * (w["Intentional"] / total_other) | |
| w["Conventional"] -= diff * (w["Conventional"] / total_other) | |
| w["Arbitrary"] = a_new | |
| st.session_state.prev_weights = w.copy() | |
| st.rerun() | |
| # --- Triangle vertices --- | |
| vertices = np.array([ | |
| [0.5, np.sqrt(3)/2], # Intentional | |
| [0, 0], # Conventional | |
| [1, 0] # Arbitrary | |
| ]) | |
| # Point from barycentric coords | |
| point = ( | |
| w["Intentional"] * vertices[0] + | |
| w["Conventional"] * vertices[1] + | |
| w["Arbitrary"] * vertices[2] | |
| ) | |
| # --- Plot --- | |
| fig, ax = plt.subplots() | |
| ax.plot(*np.append(vertices, [vertices[0]], axis=0).T) | |
| ax.text(*vertices[0], "Intentional", ha="center", va="bottom", color="green", weight="heavy") | |
| ax.text(*vertices[1], "Conventional", ha="right", va="top", color="green", weight="heavy") | |
| ax.text(*vertices[2], "Arbitrary", ha="left", va="top", color="green", weight="heavy") | |
| ax.scatter(point[0], point[1], c="white", s=10000) | |
| ax.scatter(point[0], point[1], c="orange", s=10000, zorder=5, alpha=0.3) | |
| ax.set_aspect("equal") | |
| ax.axis("off") | |
| fig.patch.set_alpha(0) | |
| ax.patch.set_alpha(0) | |
| # --- Dummy points --- | |
| locations = [ | |
| (0.9, 0.1, "Random Seeds", "Randomness is highly arbitrary, without any convention or intentionality.", | |
| "left", "bottom"), | |
| (0.35, 0.06, "Neural networks for Tabular Data", "Using an unnecessarily complex AI model for a setting where its not needed is highly conventional, a bit arbitrary, and has very low intentionality.", | |
| "left", "bottom"), | |
| (0.4, 0.5, "Foundation Model for a Complex Task", "Using a language model for a complex task is intentional, however, it also has conventionality to it, as a specialized model could have worked. Low arbitrariness.", | |
| "center", "bottom"), | |
| (0.5, 0.7, "Best Bias Mitigation for a Particular Setup", "Choosing the most appropriate bias mitigation technique, specialized for the particular setup, is highly intentional.", | |
| "center", "bottom"), | |
| (0.7, 0.5, "Randomly chosen Regularization", "Adding regularization, but choosing the regularization method randomly, creates a decision that is intentional and arbitrary, while avoiding conventionality.", | |
| "center", "top"), | |
| (0.15, 0.15, "ReLU Activation as Default", "Choosing some popular architecture component without testing what other components could have worked, is a highly conventional decision.", | |
| "center", "bottom"), | |
| ] | |
| torch_radius = 0.177 | |
| explanations = [] | |
| for (x, y, label, labeltext, ha, va) in locations: | |
| dist = np.linalg.norm([x - point[0], y - point[1]]) | |
| if dist <= torch_radius: | |
| ax.scatter(x, y, c="red", s=50, zorder=6) | |
| if va=="bottom": | |
| ax.text(x, y + 0.03, label, ha=ha, va=va, color="red", zorder=6, weight="heavy") | |
| elif va=="top": | |
| ax.text(x, y - 0.03, label, ha=ha, va=va, color="red", zorder=6, weight="heavy") | |
| explanations.append((label, labeltext)) | |
| else: | |
| ax.scatter(x, y, c="red", s=50, zorder=6, alpha=0.3) | |
| col1, col2, col3 = st.columns([0.3, 1, 0.3]) | |
| with col2: | |
| st.pyplot(fig) | |
| if len(explanations) > 0: | |
| text_to_show = "" | |
| for label, labeltext in explanations: | |
| text_to_show += "<b>" + label + ":</b> " + labeltext + "<br><br>" | |
| add_red_text(text_to_show) | |