cyberosa
fixing error in one graph
3546de0
import pandas as pd
import gradio as gr
import matplotlib.pyplot as plt
import seaborn as sns
from typing import Tuple
import plotly.express as px
import numpy as np
VOLUME_FACTOR_REGULARIZATION = 0.25
UNSCALED_WEIGHTED_ACCURACY_INTERVAL = (-0.5, 80.5)
SCALED_WEIGHTED_ACCURACY_INTERVAL = (0, 1)
# tools palette as dictionary
tools_palette = {
"prediction-request-reasoning": "darkorchid",
"claude-prediction-offline": "rebeccapurple",
"prediction-request-reasoning-claude": "slateblue",
"prediction-request-rag-claude": "steelblue",
"prediction-online": "darkcyan",
"prediction-offline": "mediumaquamarine",
"claude-prediction-online": "mediumseagreen",
"prediction-online-sme": "yellowgreen",
"prediction-url-cot-claude": "gold",
"prediction-offline-sme": "orange",
"prediction-request-rag": "chocolate",
}
HEIGHT = 400
WIDTH = 1100
def scale_value(
value: float,
min_max_bounds: Tuple[float, float],
scale_bounds: Tuple[float, float] = (0, 1),
) -> float:
"""Perform min-max scaling on a value."""
min_, max_ = min_max_bounds
current_range = max_ - min_
# normalize between 0-1
std = (value - min_) / current_range
# scale between min_bound and max_bound
min_bound, max_bound = scale_bounds
target_range = max_bound - min_bound
return std * target_range + min_bound
def get_weighted_accuracy(row, global_requests: int):
"""Function to compute the weighted accuracy of a tool"""
return scale_value(
(
row["tool_accuracy"]
+ (row["total_requests"] / global_requests) * VOLUME_FACTOR_REGULARIZATION
),
UNSCALED_WEIGHTED_ACCURACY_INTERVAL,
SCALED_WEIGHTED_ACCURACY_INTERVAL,
)
def compute_weighted_accuracy(tools_accuracy: pd.DataFrame):
global_requests = tools_accuracy.total_requests.sum()
tools_accuracy["weighted_accuracy"] = tools_accuracy.apply(
lambda x: get_weighted_accuracy(x, global_requests), axis=1
)
return tools_accuracy
def plot_tools_accuracy_rotated_graph(tools_accuracy_info: pd.DataFrame):
tools_accuracy_info = tools_accuracy_info.sort_values(
by="tool_accuracy", ascending=False
)
fig = px.bar(
tools_accuracy_info,
x="tool",
y="tool_accuracy",
color="tool",
color_discrete_map=tools_palette,
)
fig.update_layout(
xaxis_title="Tool",
yaxis_title="Mech tool_accuracy (%)",
)
fig.update_layout(width=WIDTH, height=HEIGHT)
# fig.update_xaxes(tickangle=45)
fig.update_xaxes(showticklabels=False)
return gr.Plot(
value=fig,
)
def plot_tools_weighted_accuracy_graph(tools_accuracy_info: pd.DataFrame):
tools_accuracy_info = tools_accuracy_info.sort_values(
by="weighted_accuracy", ascending=False
)
# Create the Seaborn bar plot
# sns.set_theme(palette="viridis")
plt.figure(figsize=(25, 10))
plot = sns.barplot(
tools_accuracy_info,
x="weighted_accuracy",
y="tool",
hue="tool",
dodge=False,
palette=tools_palette,
)
plt.xlabel("Weighted accuracy metric", fontsize=20)
plt.ylabel("tool", fontsize=20)
plt.tick_params(axis="y", labelsize=12)
return gr.Plot(value=plot.get_figure())
def plot_tools_weighted_accuracy_rotated_graph(
tools_accuracy_info: pd.DataFrame,
) -> gr.Plot:
tools_accuracy_info = tools_accuracy_info.sort_values(
by="weighted_accuracy", ascending=False
)
fig = px.bar(
tools_accuracy_info,
x="tool",
y="weighted_accuracy",
color="tool",
color_discrete_map=tools_palette,
)
fig.update_layout(
xaxis_title="Tool",
yaxis_title="Weighted accuracy metric",
)
fig.update_layout(width=WIDTH, height=HEIGHT)
# fig.update_xaxes(tickangle=45)
fig.update_xaxes(showticklabels=False)
return gr.Plot(
value=fig,
)
def plot_mech_requests_topthree_tools(
daily_mech_requests: pd.DataFrame, tools_accuracy_info: pd.DataFrame, top: int
):
"""Function to plot the percentage of mech requests from the top three tools"""
# Get the top three tools
top_tools = tools_accuracy_info.sort_values(
by="tool_accuracy", ascending=False
).head(top)
top_tools = top_tools.tool.tolist()
# Filter the daily mech requests for the top three tools
daily_mech_requests_local_copy = daily_mech_requests.copy()
daily_mech_requests_local_copy = daily_mech_requests_local_copy[
daily_mech_requests_local_copy["market_creator"] == "all"
]
# Get the daily total of mech requests no matter the tool
total_daily_mech_requests = (
daily_mech_requests_local_copy.groupby(["request_date"])
.agg({"total_mech_requests": "sum"})
.reset_index()
)
print("total_daily_mech_requests", total_daily_mech_requests.head())
total_daily_mech_requests.rename(
columns={"total_mech_requests": "total_daily_mech_requests"},
inplace=True,
)
# Merge the total daily mech requests with the daily mech requests
daily_mech_requests_local_copy = pd.merge(
daily_mech_requests_local_copy,
total_daily_mech_requests,
on="request_date",
how="left",
)
# Compute the percentage of mech requests for each tool
daily_mech_requests_local_copy["percentage"] = (
daily_mech_requests_local_copy["total_mech_requests"]
/ daily_mech_requests_local_copy["total_daily_mech_requests"]
) * 100
daily_mech_requests_local_copy = daily_mech_requests_local_copy[
daily_mech_requests_local_copy.tool.isin(top_tools)
]
# remove the market_creator column
daily_mech_requests_local_copy = daily_mech_requests_local_copy.drop(
columns=["market_creator"]
)
# Create a pivot table to get the total mech requests per tool
pivoted = daily_mech_requests_local_copy.pivot(
index="request_date", columns="tool", values="percentage"
)
# Sort the columns for each row independently
sorted_values = np.sort(pivoted.values, axis=1)[
:, ::-1
] # sort and reverse (descending)
sorted_columns = np.argsort(pivoted.values, axis=1)[:, ::-1] # get sorting indices
sorted_df = pd.DataFrame(
sorted_values,
index=pivoted.index,
columns=[
pivoted.columns[i] for i in sorted_columns[0]
], # use first row's order
)
sorted_long = sorted_df.reset_index().melt(
id_vars=["request_date"], var_name="tool", value_name="percentage"
)
fig = px.bar(
sorted_long,
x="request_date",
y="percentage",
color="tool",
color_discrete_map=tools_palette,
)
fig.update_layout(
xaxis_title="Day of the request",
yaxis_title="Percentage of Total daily mech requests",
legend=dict(yanchor="top", y=0.5),
)
fig.update_layout(width=WIDTH, height=HEIGHT)
return gr.Plot(value=fig)