# AUTOGENERATED! DO NOT EDIT! File to edit: 03a_image_archs.ipynb.

# %% auto 0
__all__ = ['result_options', 'activity_options', 'col_options', 'subs', 'is_fullmatch', 'drop_tf', 'result_option',
           'activity_option', 'col_option', 'size_col_option', 'title_dict', 'df', 'fig', 'col1', 'col2', 'get_results',
           'get_integrated_data', 'get_filtered_data', 'get_data', 'plot_selection']

# %% 03a_image_archs.ipynb 11
import pandas as pd
import plotly.express as px
from fastcore.all import *
import streamlit as st

# %% 03a_image_archs.ipynb 12
st.set_page_config(page_title="Which Image Model is best?",layout="wide")

# %% 03a_image_archs.ipynb 14
def get_results(result_option = 'original'):
  suffix = "" if result_option == 'original' else "-real"
  url_results = f"https://github.com/huggingface/pytorch-image-models/raw/main/results/results-imagenet{suffix}.csv"
  df_results = pd.read_csv(url_results); df_results.head()
  df_results['model_org'] = df_results['model']
  df_results['model'] = df_results['model'].str.split('.').str[0]
  return df_results

# %% 03a_image_archs.ipynb 16
def get_integrated_data(activity_option, result_option):
  df_results = get_results(result_option)
  url_benchmark = f"https://github.com/huggingface/pytorch-image-models/raw/main/results/benchmark-{activity_option}-amp-nhwc-pt112-cu113-rtx3090.csv"
  df_benchmark = pd.read_csv(url_benchmark)
  df_integrated = df_results.merge(df_benchmark, on='model')
  df_integrated['is_tensorflow_model'] = df_integrated.model.str.split('_').str[0] =='tf'
  df_integrated['family'] = df_integrated.model.str.removeprefix("tf_").str.removeprefix("legacy_").str.removeprefix("nf_").str.removeprefix("nf_").str.extract('^([a-z]+?(?:v2|v3)?)(?:\d|_|$)')[0].values
  df_integrated.loc[df_integrated.model.str.contains('in22'), 'family'] = df_integrated.loc[df_integrated.model.str.contains('in22'), 'family'] + "_in22"
  df_integrated.loc[df_integrated.model.str.contains('resnet.*d'), 'family'] = df_integrated.loc[df_integrated.model.str.contains('resnet.*d'), 'family'] + "d"
  return df_integrated[~df_integrated.model.str.endswith('gn')] # Group norm models. Why Jeremy eliminated them from analysis?

# %% 03a_image_archs.ipynb 18
@st.cache_data
def get_filtered_data(df_integrated, subs, is_fullmatch=False, drop_tf=True):
  if drop_tf: df_integrated = df_integrated[~df_integrated.is_tensorflow_model]
  if not subs: return df_integrated
  elif is_fullmatch: return df_integrated[df_integrated.family.str.fullmatch(subs)]
  else: return df_integrated[df_integrated.model.str.contains(subs)]

# %% 03a_image_archs.ipynb 19
def get_data(col_option, activity_option, result_option, subs, is_fullmatch=False, drop_tf=True):
  col = "_".join([activity_option, col_option])
  df_integrated = get_integrated_data(activity_option, result_option)
  df_integrated = get_filtered_data(df_integrated, subs, is_fullmatch=is_fullmatch, drop_tf=drop_tf)
  df_integrated['secs'] =1./df_integrated[col]
  return df_integrated

# %% 03a_image_archs.ipynb 20
def plot_selection(df, title, col_option, activity_option, w=1000, h=800):
  size_col = "_".join([activity_option, col_option])
  return px.scatter(df, width=w, height=h, size=df[size_col]**2,trendline="ols", trendline_options={'log_x':True},
                    title=title, x="secs",log_x=True, y='top1', log_y=True,
                    color="family", hover_name='model_org',
                    hover_data=[size_col])


# %% 03a_image_archs.ipynb 21
result_options = ['original', 'real'] #result = 'real'
activity_options = ['train', 'infer']
col_options = ['samples_per_sec', 'step_time', 'batch_size', 'img_size', 'gmacs', 'macts']
subs = '^re[sg]netd?|beit|convnext|levit|efficient|vit|vgg|swin'
is_fullmatch = False
drop_tf = False
subs = 'levit|resnetd?|regnetx|vgg|convnext.*|efficientnetv2|beit|swin'
is_fullmatch = True
result_option = result_options[0]
activity_option = activity_options[1]
col_option = col_options[0]
size_col_option = col_options[3]
title_dict = dict(zip(activity_options, ['Training', "Inference"]))
df = get_data(col_option, activity_option, result_option, subs, is_fullmatch=is_fullmatch, drop_tf=drop_tf)
fig = plot_selection(df, title_dict[activity_option], size_col_option, activity_option)

# %% 03a_image_archs.ipynb 25
st.title("Which Image Model is best?")
col1, col2 = st.columns([1,3])
with col1:
    st.header("Settings")
    result_option = st.selectbox("Please choose dataset", result_options)
    activity_option = st.selectbox("Please choose activity", activity_options)
    subs = st.text_input("Subs", value='levit|resnetd?|regnetx|vgg|convnext.*|efficientnetv2|beit|swin')
    is_fullmatch = st.checkbox("Is fullmatch", value=True)
    drop_tf = st.checkbox("Drop Tensorflow Models", value=False)
    col_option = st.selectbox("Please choose col_option", col_options)
    size_col_option = st.selectbox("Please choose sizing col_option", col_options, index=3)
with col2:
    title_dict = dict(zip(activity_options, ['Training', "Inference"]))
    df = get_data(col_option, activity_option, result_option, subs, is_fullmatch=is_fullmatch, drop_tf=drop_tf)
    fig = plot_selection(df, None, size_col_option, activity_option, h=500, w=1000)
    # Plot!
    st.header(title_dict[activity_option])
    st.plotly_chart(fig, use_container_width=True, height=500)