app.py

# --- START OF MODIFIED FILE app.py ---

import gradio as gr
import pandas as pd
import plotly.express as px
import time
from datasets import load_dataset # Import the datasets library

# --- Constants ---
# REMOVED the old MODEL_SIZE_RANGES dictionary.
# NEW: Define the discrete steps for the parameter range slider.
PARAM_CHOICES = ['< 1B', '1B', '5B', '12B', '32B', '64B', '128B', '256B', '> 500B']
PARAM_CHOICES_DEFAULT = [PARAM_CHOICES[0], PARAM_CHOICES[-1]]

# The Hugging Face dataset ID to load.
HF_DATASET_ID = "evijit/orgstats_daily_data"

TAG_FILTER_CHOICES = [
    "Audio & Speech", "Time series", "Robotics", "Music", "Video", "Images",
    "Text", "Biomedical", "Sciences"
]

PIPELINE_TAGS = [
 'text-generation', 'text-to-image', 'text-classification', 'text2text-generation',
 'audio-to-audio', 'feature-extraction', 'image-classification', 'translation',
 'reinforcement-learning', 'fill-mask', 'text-to-speech', 'automatic-speech-recognition',
 'image-text-to-text', 'token-classification', 'sentence-similarity', 'question-answering',
 'image-feature-extraction', 'summarization', 'zero-shot-image-classification',
 'object-detection', 'image-segmentation', 'image-to-image', 'image-to-text',
 'audio-classification', 'visual-question-answering', 'text-to-video',
 'zero-shot-classification', 'depth-estimation', 'text-ranking', 'image-to-video',
 'multiple-choice', 'unconditional-image-generation', 'video-classification',
 'text-to-audio', 'time-series-forecasting', 'any-to-any', 'video-text-to-text',
 'table-question-answering',
]


def load_models_data():
    """
    Loads the pre-processed models data using the HF datasets library.
    """
    overall_start_time = time.time()
    print(f"Attempting to load dataset from Hugging Face Hub: {HF_DATASET_ID}")

    expected_cols = [
        'id', 'downloads', 'downloadsAllTime', 'likes', 'pipeline_tag', 'tags', 'params',
        'organization', 'has_audio', 'has_speech', 'has_music',
        'has_robot', 'has_bio', 'has_med', 'has_series', 'has_video', 'has_image',
        'has_text', 'has_science', 'is_audio_speech', 'is_biomed',
        'data_download_timestamp'
    ]

    try:
        dataset_dict = load_dataset(HF_DATASET_ID)
        
        if not dataset_dict:
             raise ValueError(f"Dataset '{HF_DATASET_ID}' loaded but appears empty.")

        split_name = list(dataset_dict.keys())[0]
        print(f"Using dataset split: '{split_name}'. Converting to Pandas.")
        df = dataset_dict[split_name].to_pandas()
        elapsed = time.time() - overall_start_time
        
        missing_cols = [col for col in expected_cols if col not in df.columns]
        if missing_cols:
            # The 'params' column is crucial for the new slider.
            if 'params' in missing_cols:
                raise ValueError(f"FATAL: Loaded dataset is missing the crucial 'params' column.")
            print(f"Warning: Loaded dataset is missing some expected columns: {missing_cols}.")

        # Ensure 'params' column is numeric, coercing errors to NaN and then filling with 0.
        # This is important for filtering. Assumes params are in billions.
        if 'params' in df.columns:
            df['params'] = pd.to_numeric(df['params'], errors='coerce').fillna(0)
        else:
            # If 'params' is missing after all, create a dummy column to prevent crashes.
            df['params'] = 0
            print("CRITICAL WARNING: 'params' column not found in data. Parameter filtering will not work.")

        msg = f"Successfully loaded dataset '{HF_DATASET_ID}' (split: {split_name}) from HF Hub in {elapsed:.2f}s. Shape: {df.shape}"
        print(msg)
        return df, True, msg
        
    except Exception as e:
        err_msg = f"Failed to load dataset '{HF_DATASET_ID}' from Hugging Face Hub. Error: {e}"
        print(err_msg)
        return pd.DataFrame(), False, err_msg

# --- NEW: Helper function to parse slider labels into numerical values ---
def get_param_range_values(param_range_labels):
    """Converts a list of two string labels from the slider into a numerical min/max tuple."""
    if not param_range_labels or len(param_range_labels) != 2:
        return None, None
        
    min_label, max_label = param_range_labels
    
    # Min value logic: '< 1B' becomes 0, otherwise parse the number.
    min_val = 0.0 if '<' in min_label else float(min_label.replace('B', ''))
    
    # Max value logic: '> 500B' becomes infinity, otherwise parse the number.
    max_val = float('inf') if '>' in max_label else float(max_label.replace('B', ''))
    
    return min_val, max_val

# --- MODIFIED: Function signature and filtering logic updated for parameter range ---
def make_treemap_data(df, count_by, top_k=25, tag_filter=None, pipeline_filter=None, param_range=None, skip_orgs=None):
    if df is None or df.empty: return pd.DataFrame()
    filtered_df = df.copy()
    col_map = { "Audio & Speech": "is_audio_speech", "Music": "has_music", "Robotics": "has_robot",
                "Biomedical": "is_biomed", "Time series": "has_series", "Sciences": "has_science",
                "Video": "has_video", "Images": "has_image", "Text": "has_text"}
    
    if tag_filter and tag_filter in col_map:
        target_col = col_map[tag_filter]
        if target_col in filtered_df.columns:
            filtered_df = filtered_df[filtered_df[target_col]]
        else:
            print(f"Warning: Tag filter column '{col_map[tag_filter]}' not found in DataFrame.")

    if pipeline_filter:
        if "pipeline_tag" in filtered_df.columns:
            filtered_df = filtered_df[filtered_df["pipeline_tag"].astype(str) == pipeline_filter]
        else:
            print(f"Warning: 'pipeline_tag' column not found for filtering.")

    # --- MODIFIED: Filtering logic now uses the numerical parameter range ---
    if param_range:
        min_params, max_params = get_param_range_values(param_range)
        is_default_range = (param_range == PARAM_CHOICES_DEFAULT)

        # Only filter if the range is not the default full range
        if not is_default_range and 'params' in filtered_df.columns:
            # The 'params' column is in billions, so the values match our slider
            if min_params is not None:
                filtered_df = filtered_df[filtered_df['params'] >= min_params]
            if max_params is not None and max_params != float('inf'):
                # The upper bound is exclusive, e.g., 5B to 64B is [5, 64)
                filtered_df = filtered_df[filtered_df['params'] < max_params]
        elif 'params' not in filtered_df.columns:
             print("Warning: 'params' column not found for filtering.")


    if skip_orgs and len(skip_orgs) > 0:
        if "organization" in filtered_df.columns:
            filtered_df = filtered_df[~filtered_df["organization"].isin(skip_orgs)]
        else:
            print("Warning: 'organization' column not found for filtering.")

    if filtered_df.empty: return pd.DataFrame()

    if count_by not in filtered_df.columns:
         print(f"Warning: Metric column '{count_by}' not found. Using 0.")
         filtered_df[count_by] = 0.0
    filtered_df[count_by] = pd.to_numeric(filtered_df[count_by], errors="coerce").fillna(0.0)
    
    org_totals = filtered_df.groupby("organization")[count_by].sum().nlargest(top_k, keep='first')
    top_orgs_list = org_totals.index.tolist()
    
    treemap_data = filtered_df[filtered_df["organization"].isin(top_orgs_list)][["id", "organization", count_by]].copy()
    treemap_data["root"] = "models"
    treemap_data[count_by] = pd.to_numeric(treemap_data[count_by], errors="coerce").fillna(0.0) 
    return treemap_data

def create_treemap(treemap_data, count_by, title=None):
    if treemap_data.empty:
        fig = px.treemap(names=["No data matches filters"], parents=[""], values=[1])
        fig.update_layout(title="No data matches the selected filters", margin=dict(t=50, l=25, r=25, b=25))
        return fig
    fig = px.treemap(
        treemap_data, path=["root", "organization", "id"], values=count_by,
        title=title or f"HuggingFace Models - {count_by.capitalize()} by Organization",
        color_discrete_sequence=px.colors.qualitative.Plotly
    )
    fig.update_layout(margin=dict(t=50, l=25, r=25, b=25))
    fig.update_traces(textinfo="label+value+percent root", hovertemplate="<b>%{label}</b><br>%{value:,} " + count_by + "<br>%{percentRoot:.2%} of total<extra></extra>")
    return fig

with gr.Blocks(title="HuggingFace Model Explorer", fill_width=True) as demo:
    models_data_state = gr.State(pd.DataFrame())
    loading_complete_state = gr.State(False)

    with gr.Row(): gr.Markdown("# HuggingFace Models TreeMap Visualization")
    with gr.Row():
        with gr.Column(scale=1):
            count_by_dropdown = gr.Dropdown(label="Metric", choices=[("Downloads (last 30 days)", "downloads"), ("Downloads (All Time)", "downloadsAllTime"), ("Likes", "likes")], value="downloads")
            filter_choice_radio = gr.Radio(label="Filter Type", choices=["None", "Tag Filter", "Pipeline Filter"], value="None")
            tag_filter_dropdown = gr.Dropdown(label="Select Tag", choices=TAG_FILTER_CHOICES, value=None, visible=False)
            pipeline_filter_dropdown = gr.Dropdown(label="Select Pipeline Tag", choices=PIPELINE_TAGS, value=None, visible=False)
            
            # --- MODIFIED: Replaced Dropdown with RangeSlider and a Reset Button ---
            with gr.Group():
                with gr.Row():
                    gr.Markdown("<div style='padding-top: 10px; font-weight: 500;'>Parameters</div>")
                    reset_params_button = gr.Button("🔄 Reset", visible=False, size="sm", min_width=80)
                param_range_slider = gr.RangeSlider(
                    label=None, # Label is handled by Markdown above
                    choices=PARAM_CHOICES,
                    value=PARAM_CHOICES_DEFAULT,
                )
            # --- END OF MODIFICATION ---
            
            top_k_slider = gr.Slider(label="Number of Top Organizations", minimum=5, maximum=50, value=25, step=5)
            skip_orgs_textbox = gr.Textbox(label="Organizations to Skip (comma-separated)", value="TheBloke,MaziyarPanahi,unsloth,modularai,Gensyn,bartowski")
            generate_plot_button = gr.Button(value="Generate Plot", variant="primary", interactive=False)

        with gr.Column(scale=3):
            plot_output = gr.Plot()
            status_message_md = gr.Markdown("Initializing...")
            data_info_md = gr.Markdown("")

    # --- NEW: Event handlers for the new parameter slider and reset button ---
    def _update_reset_button_visibility(current_range):
        """Shows the reset button only if the slider is not at its default full range."""
        is_default = (current_range == PARAM_CHOICES_DEFAULT)
        return gr.update(visible=not is_default)

    def _reset_param_slider_and_button():
        """Resets the slider to its default value and hides the reset button."""
        return gr.update(value=PARAM_CHOICES_DEFAULT), gr.update(visible=False)

    param_range_slider.release(fn=_update_reset_button_visibility, inputs=param_range_slider, outputs=reset_params_button)
    reset_params_button.click(fn=_reset_param_slider_and_button, outputs=[param_range_slider, reset_params_button])
    # --- END OF NEW EVENT HANDLERS ---

    def _update_button_interactivity(is_loaded_flag):
        return gr.update(interactive=is_loaded_flag)
    loading_complete_state.change(fn=_update_button_interactivity, inputs=loading_complete_state, outputs=generate_plot_button)

    def _toggle_filters_visibility(choice):
        return gr.update(visible=choice == "Tag Filter"), gr.update(visible=choice == "Pipeline Filter")
    filter_choice_radio.change(fn=_toggle_filters_visibility, inputs=filter_choice_radio, outputs=[tag_filter_dropdown, pipeline_filter_dropdown])

    def ui_load_data_controller(progress=gr.Progress()):
        progress(0, desc=f"Loading dataset '{HF_DATASET_ID}' from Hugging Face Hub...")
        print("ui_load_data_controller called.")
        status_msg_ui = "Loading data..."
        data_info_text = ""
        current_df = pd.DataFrame()
        load_success_flag = False
        try:
            current_df, load_success_flag, status_msg_from_load = load_models_data()
            if load_success_flag:
                progress(0.9, desc="Processing loaded data...")
                if 'data_download_timestamp' in current_df.columns and not current_df.empty and pd.notna(current_df['data_download_timestamp'].iloc[0]):
                    timestamp_from_parquet = pd.to_datetime(current_df['data_download_timestamp'].iloc[0]).tz_localize('UTC')
                    data_as_of_date_display = timestamp_from_parquet.strftime('%B %d, %Y, %H:%M:%S %Z')
                else:
                    data_as_of_date_display = "Pre-processed (date unavailable)"
                
                # --- MODIFIED: Removed the old size category distribution text ---
                param_count = (current_df['params'] > 0).sum() if 'params' in current_df.columns else 0
                data_info_text = (f"### Data Information\n"
                                  f"- Source: `{HF_DATASET_ID}`\n"
                                  f"- Overall Status: {status_msg_from_load}\n" 
                                  f"- Total models loaded: {len(current_df):,}\n"
                                  f"- Models with parameter counts: {param_count:,}\n"
                                  f"- Data as of: {data_as_of_date_display}\n")
                
                status_msg_ui = "Data loaded successfully. Ready to generate plot."
            else: 
                data_info_text = f"### Data Load Failed\n- {status_msg_from_load}"
                status_msg_ui = status_msg_from_load 
        except Exception as e:
            status_msg_ui = f"An unexpected error occurred in ui_load_data_controller: {str(e)}"
            data_info_text = f"### Critical Error\n- {status_msg_ui}"
            print(f"Critical error in ui_load_data_controller: {e}")
            load_success_flag = False 
        return current_df, load_success_flag, data_info_text, status_msg_ui

    # --- MODIFIED: Updated controller signature and logic to handle new slider ---
    def ui_generate_plot_controller(metric_choice, filter_type, tag_choice, pipeline_choice, 
                                   param_range_choice, k_orgs, skip_orgs_input, df_current_models, progress=gr.Progress()):
        if df_current_models is None or df_current_models.empty:
            empty_fig = create_treemap(pd.DataFrame(), metric_choice, "Error: Model Data Not Loaded")
            error_msg = "Model data is not loaded or is empty. Please wait for data to load."
            gr.Warning(error_msg)
            return empty_fig, error_msg
        
        progress(0.1, desc="Preparing data for visualization...")
        
        tag_to_use = tag_choice if filter_type == "Tag Filter" else None
        pipeline_to_use = pipeline_choice if filter_type == "Pipeline Filter" else None
        orgs_to_skip = [org.strip() for org in skip_orgs_input.split(',') if org.strip()] if skip_orgs_input else []
        
        # Pass the param_range_choice directly to make_treemap_data
        treemap_df = make_treemap_data(df_current_models, metric_choice, k_orgs, tag_to_use, pipeline_to_use, param_range_choice, orgs_to_skip)
        
        progress(0.7, desc="Generating Plotly visualization...")

        title_labels = {"downloads": "Downloads (last 30 days)", "downloadsAllTime": "Downloads (All Time)", "likes": "Likes"}
        chart_title = f"HuggingFace Models - {title_labels.get(metric_choice, metric_choice)} by Organization"
        plotly_fig = create_treemap(treemap_df, metric_choice, chart_title)
        
        if treemap_df.empty:
            plot_stats_md = "No data matches the selected filters. Try adjusting your filters."
        else:
            total_items_in_plot = len(treemap_df['id'].unique())
            total_value_in_plot = treemap_df[metric_choice].sum()
            plot_stats_md = (f"## Plot Statistics\n- **Models shown**: {total_items_in_plot:,}\n- **Total {metric_choice}**: {int(total_value_in_plot):,}")
        
        return plotly_fig, plot_stats_md

    demo.load(
        fn=ui_load_data_controller,
        inputs=[],
        outputs=[models_data_state, loading_complete_state, data_info_md, status_message_md]
    )

    # --- MODIFIED: Updated the inputs list for the click event ---
    generate_plot_button.click(
        fn=ui_generate_plot_controller,
        inputs=[count_by_dropdown, filter_choice_radio, tag_filter_dropdown, pipeline_filter_dropdown,
                param_range_slider, top_k_slider, skip_orgs_textbox, models_data_state],
        outputs=[plot_output, status_message_md]
    )

if __name__ == "__main__":
    print(f"Application starting. Data will be loaded from Hugging Face dataset: {HF_DATASET_ID}")
    demo.queue().launch()

# --- END OF MODIFIED FILE app.py ---