1

README.md

@@ -1,20 +1,29 @@
 ---
-title: Humanlike
+title: HHEM Leaderboard
 emoji: 🥇
 colorFrom: green
 colorTo: indigo
 sdk: gradio
+sdk_version: 4.37.1
 app_file: app.py
 pinned: true
-license: cc-by-nd-4.0
+license: apache-2.0
+tags:
+- leaderboard
 ---
 
-# Start the configuration
 
-Most of the variables to change for a default leaderboard are in `src/env.py` (replace the path for your leaderboard) and `src/about.py` (for tasks).
+python>3.10
+pip spacy
+python -m spacy download en_core_web_sm
+pip install google.generativeai
 
-Results files should have the following format and be stored as json files:
-```json
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+Most of the variables to change for a default leaderboard are in env (replace the path for your leaderboard) and src/display/about.
+
+Results files should have the following format:
+```
 {
     "config": {
         "model_dtype": "torch.float16", # or torch.bfloat16 or 8bit or 4bit
@@ -32,13 +41,4 @@ Results files should have the following format and be stored as json files:
 }
 ```
 
-Request files are created automatically by this tool.
-
-If you encounter problem on the space, don't hesitate to restart it to remove the create eval-queue, eval-queue-bk, eval-results and eval-results-bk created folder.
-
-# Code logic for more complex edits
-
-You'll find 
-- the main table' columns names and properties in `src/display/utils.py`
-- the logic to read all results and request files, then convert them in dataframe lines, in `src/leaderboard/read_evals.py`, and `src/populate.py`
-- teh logic to allow or filter submissions in `src/submission/submit.py` and `src/submission/check_validity.py`
+Request files are created automatically by this tool.

app.py

@@ -1,110 +1,234 @@
 import gradio as gr
-from gradio_leaderboard import Leaderboard, ColumnFilter, SelectColumns
 import pandas as pd
 from apscheduler.schedulers.background import BackgroundScheduler
 from huggingface_hub import snapshot_download
 
-from src.about import (
-    CITATION_BUTTON_LABEL,
-    CITATION_BUTTON_TEXT,
-    EVALUATION_QUEUE_TEXT,
-    INTRODUCTION_TEXT,
-    LLM_BENCHMARKS_TEXT,
-    TITLE,
-)
+import src.display.about as about
 from src.display.css_html_js import custom_css
-from src.display.utils import (
-    BENCHMARK_COLS,
-    COLS,
-    EVAL_COLS,
-    EVAL_TYPES,
-    AutoEvalColumn,
-    ModelType,
-    fields,
-    WeightType,
-    Precision
-)
-from src.envs import API, EVAL_REQUESTS_PATH, EVAL_RESULTS_PATH, QUEUE_REPO, REPO_ID, RESULTS_REPO, TOKEN
-from src.populate import get_evaluation_queue_df, get_leaderboard_df
-from src.submission.submit import add_new_eval
+import src.display.utils as utils
+import src.envs as envs
+import src.populate as populate
+import src.submission.submit as submit
 
 
 def restart_space():
-    API.restart_space(repo_id=REPO_ID)
+    envs.API.restart_space(repo_id=envs.REPO_ID, token=envs.TOKEN)
 
-### Space initialisation
 try:
-    print(EVAL_REQUESTS_PATH)
+    print(envs.EVAL_REQUESTS_PATH)
     snapshot_download(
-        repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
+        repo_id=envs.QUEUE_REPO, local_dir=envs.EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30
     )
 except Exception:
     restart_space()
 try:
-    print(EVAL_RESULTS_PATH)
+    print(envs.EVAL_RESULTS_PATH)
     snapshot_download(
-        repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30, token=TOKEN
+        repo_id=envs.RESULTS_REPO, local_dir=envs.EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30
     )
 except Exception:
     restart_space()
 
-
-LEADERBOARD_DF = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
+raw_data, original_df = populate.get_leaderboard_df(envs.EVAL_RESULTS_PATH, envs.EVAL_REQUESTS_PATH, utils.COLS, utils.BENCHMARK_COLS)
+leaderboard_df = original_df.copy()
 
 (
     finished_eval_queue_df,
     running_eval_queue_df,
     pending_eval_queue_df,
-) = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
-
-def init_leaderboard(dataframe):
-    if dataframe is None or dataframe.empty:
-        raise ValueError("Leaderboard DataFrame is empty or None.")
-    return Leaderboard(
-        value=dataframe,
-        datatype=[c.type for c in fields(AutoEvalColumn)],
-        select_columns=SelectColumns(
-            default_selection=[c.name for c in fields(AutoEvalColumn) if c.displayed_by_default],
-            cant_deselect=[c.name for c in fields(AutoEvalColumn) if c.never_hidden],
-            label="Select Columns to Display:",
-        ),
-        search_columns=[AutoEvalColumn.model.name, AutoEvalColumn.license.name],
-        hide_columns=[c.name for c in fields(AutoEvalColumn) if c.hidden],
-        filter_columns=[
-            ColumnFilter(AutoEvalColumn.model_type.name, type="checkboxgroup", label="Model types"),
-            ColumnFilter(AutoEvalColumn.precision.name, type="checkboxgroup", label="Precision"),
-            ColumnFilter(
-                AutoEvalColumn.params.name,
-                type="slider",
-                min=0.01,
-                max=150,
-                label="Select the number of parameters (B)",
-            ),
-            ColumnFilter(
-                AutoEvalColumn.still_on_hub.name, type="boolean", label="Deleted/incomplete", default=True
-            ),
-        ],
-        bool_checkboxgroup_label="Hide models",
-        interactive=False,
-    )
+) = populate.get_evaluation_queue_df(envs.EVAL_REQUESTS_PATH, utils.EVAL_COLS)
+
+
+# Searching and filtering
+def update_table(
+    hidden_df: pd.DataFrame,
+    columns: list,
+    type_query: list,
+    precision_query: str,
+    size_query: list,
+    show_deleted: bool,
+    query: str,
+):
+    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted)
+    filtered_df = filter_queries(query, filtered_df)
+    df = select_columns(filtered_df, columns)
+    return df
+
+
+def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
+    return df[(df[utils.AutoEvalColumn.dummy.name].str.contains(query, case=False))]
+
+
+def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
+    always_here_cols = [
+        utils.AutoEvalColumn.model_type_symbol.name,
+        utils.AutoEvalColumn.model.name,
+    ]
+    # We use COLS to maintain sorting
+    filtered_df = df[
+        always_here_cols + [c for c in utils.COLS if c in df.columns and c in columns] + [utils.AutoEvalColumn.dummy.name]
+    ]
+    return filtered_df
+
+
+def filter_queries(query: str, filtered_df: pd.DataFrame) -> pd.DataFrame:
+    final_df = []
+    if query != "":
+        queries = [q.strip() for q in query.split(";")]
+        for _q in queries:
+            _q = _q.strip()
+            if _q != "":
+                temp_filtered_df = search_table(filtered_df, _q)
+                if len(temp_filtered_df) > 0:
+                    final_df.append(temp_filtered_df)
+        if len(final_df) > 0:
+            filtered_df = pd.concat(final_df)
+            filtered_df = filtered_df.drop_duplicates(
+                subset=[utils.AutoEvalColumn.model.name, utils.AutoEvalColumn.precision.name, utils.AutoEvalColumn.revision.name]
+            )
+
+    return filtered_df
+
+
+def filter_models(
+    df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool
+) -> pd.DataFrame:
+    # Show all models
+    # if show_deleted:
+    #   filtered_df = df
+    # else:  # Show only still on the hub models
+        # filtered_df = df[df[utils.AutoEvalColumn.still_on_hub.name]]
+
+    filtered_df = df
+    
+    type_emoji = [t[0] for t in type_query]
+    filtered_df = filtered_df.loc[df[utils.AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
+    filtered_df = filtered_df.loc[df[utils.AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
+
+    numeric_interval = pd.IntervalIndex(sorted([utils.NUMERIC_INTERVALS[s] for s in size_query]))
+    params_column = pd.to_numeric(df[utils.AutoEvalColumn.params.name], errors="coerce")
+    mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
+    filtered_df = filtered_df.loc[mask]
+
+    return filtered_df
 
 
 demo = gr.Blocks(css=custom_css)
 with demo:
-    gr.HTML(TITLE)
-    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
+    gr.HTML(about.TITLE)
+    gr.Markdown(about.INTRODUCTION_TEXT, elem_classes="markdown-text")
 
     with gr.Tabs(elem_classes="tab-buttons") as tabs:
         with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):
-            leaderboard = init_leaderboard(LEADERBOARD_DF)
+            with gr.Row():
+                with gr.Column():
+                    with gr.Row():
+                        search_bar = gr.Textbox(
+                            placeholder=" 🔍 Search for your model (separate multiple queries with `;`) and press ENTER...",
+                            show_label=False,
+                            elem_id="search-bar",
+                        )
+                    with gr.Row():
+                        shown_columns = gr.CheckboxGroup(
+                            choices=[
+                                c.name
+                                for c in utils.fields(utils.AutoEvalColumn)
+                                if not c.hidden and not c.never_hidden and not c.dummy
+                            ],
+                            value=[
+                                c.name
+                                for c in utils.fields(utils.AutoEvalColumn)
+                                if c.displayed_by_default and not c.hidden and not c.never_hidden
+                            ],
+                            label="Select columns to show",
+                            elem_id="column-select",
+                            interactive=True,
+                        )
+                    with gr.Row():
+                        deleted_models_visibility = gr.Checkbox(
+                            value=False, label="Show gated/private/deleted models", interactive=True
+                        )
+                with gr.Column(min_width=320):
+                    #with gr.Box(elem_id="box-filter"):
+                    filter_columns_type = gr.CheckboxGroup(
+                        label="Model types",
+                        choices=[t.to_str() for t in utils.ModelType],
+                        value=[t.to_str() for t in utils.ModelType],
+                        interactive=True,
+                        elem_id="filter-columns-type",
+                    )
+                    filter_columns_precision = gr.CheckboxGroup(
+                        label="Precision",
+                        choices=[i.value.name for i in utils.Precision],
+                        value=[i.value.name for i in utils.Precision],
+                        interactive=True,
+                        elem_id="filter-columns-precision",
+                    )
+                    filter_columns_size = gr.CheckboxGroup(
+                        label="Model sizes (in billions of parameters)",
+                        choices=list(utils.NUMERIC_INTERVALS.keys()),
+                        value=list(utils.NUMERIC_INTERVALS.keys()),
+                        interactive=True,
+                        elem_id="filter-columns-size",
+                    )
+
+            leaderboard_table = gr.components.Dataframe(
+                value=leaderboard_df[
+                    [c.name for c in utils.fields(utils.AutoEvalColumn) if c.never_hidden]
+                    + shown_columns.value
+                    + [utils.AutoEvalColumn.dummy.name]
+                ],
+                headers=[c.name for c in utils.fields(utils.AutoEvalColumn) if c.never_hidden] + shown_columns.value,
+                datatype=utils.TYPES,
+                elem_id="leaderboard-table",
+                interactive=False,
+                visible=True,
+                column_widths=["2%", "33%"]
+            )
+
+            # Dummy leaderboard for handling the case when the user uses backspace key
+            hidden_leaderboard_table_for_search = gr.components.Dataframe(
+                value=original_df[utils.COLS],
+                headers=utils.COLS,
+                datatype=utils.TYPES,
+                visible=False,
+            )
+            search_bar.submit(
+                update_table,
+                [
+                    hidden_leaderboard_table_for_search,
+                    shown_columns,
+                    filter_columns_type,
+                    filter_columns_precision,
+                    filter_columns_size,
+                    deleted_models_visibility,
+                    search_bar,
+                ],
+                leaderboard_table,
+            )
+            for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size, deleted_models_visibility]:
+                selector.change(
+                    update_table,
+                    [
+                        hidden_leaderboard_table_for_search,
+                        shown_columns,
+                        filter_columns_type,
+                        filter_columns_precision,
+                        filter_columns_size,
+                        deleted_models_visibility,
+                        search_bar,
+                    ],
+                    leaderboard_table,
+                    queue=True,
+                )
 
         with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
-            gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
+            gr.Markdown(about.LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
 
         with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
             with gr.Column():
                 with gr.Row():
-                    gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
+                    gr.Markdown(about.EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
 
                 with gr.Column():
                     with gr.Accordion(
@@ -114,8 +238,8 @@ with demo:
                         with gr.Row():
                             finished_eval_table = gr.components.Dataframe(
                                 value=finished_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
+                                headers=utils.EVAL_COLS,
+                                datatype=utils.EVAL_TYPES,
                                 row_count=5,
                             )
                     with gr.Accordion(
@@ -125,8 +249,8 @@ with demo:
                         with gr.Row():
                             running_eval_table = gr.components.Dataframe(
                                 value=running_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
+                                headers=utils.EVAL_COLS,
+                                datatype=utils.EVAL_TYPES,
                                 row_count=5,
                             )
 
@@ -137,8 +261,8 @@ with demo:
                         with gr.Row():
                             pending_eval_table = gr.components.Dataframe(
                                 value=pending_eval_queue_df,
-                                headers=EVAL_COLS,
-                                datatype=EVAL_TYPES,
+                                headers=utils.EVAL_COLS,
+                                datatype=utils.EVAL_TYPES,
                                 row_count=5,
                             )
             with gr.Row():
@@ -149,7 +273,7 @@ with demo:
                     model_name_textbox = gr.Textbox(label="Model name")
                     revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
                     model_type = gr.Dropdown(
-                        choices=[t.to_str(" : ") for t in ModelType if t != ModelType.Unknown],
+                        choices=[t.to_str(" : ") for t in utils.ModelType if t != utils.ModelType.Unknown],
                         label="Model type",
                         multiselect=False,
                         value=None,
@@ -158,14 +282,14 @@ with demo:
 
                 with gr.Column():
                     precision = gr.Dropdown(
-                        choices=[i.value.name for i in Precision if i != Precision.Unknown],
+                        choices=[i.value.name for i in utils.Precision if i != utils.Precision.Unknown],
                         label="Precision",
                         multiselect=False,
                         value="float16",
                         interactive=True,
                     )
                     weight_type = gr.Dropdown(
-                        choices=[i.value.name for i in WeightType],
+                        choices=[i.value.name for i in utils.WeightType],
                         label="Weights type",
                         multiselect=False,
                         value="Original",
@@ -176,7 +300,7 @@ with demo:
             submit_button = gr.Button("Submit Eval")
             submission_result = gr.Markdown()
             submit_button.click(
-                add_new_eval,
+                submit.add_new_eval,
                 [
                     model_name_textbox,
                     base_model_name_textbox,
@@ -191,8 +315,8 @@ with demo:
     with gr.Row():
         with gr.Accordion("📙 Citation", open=False):
             citation_button = gr.Textbox(
-                value=CITATION_BUTTON_TEXT,
-                label=CITATION_BUTTON_LABEL,
+                value=about.CITATION_BUTTON_TEXT,
+                label=about.CITATION_BUTTON_LABEL,
                 lines=20,
                 elem_id="citation-button",
                 show_copy_button=True,
@@ -201,4 +325,4 @@ with demo:
 scheduler = BackgroundScheduler()
 scheduler.add_job(restart_space, "interval", seconds=1800)
 scheduler.start()
-demo.queue(default_concurrency_limit=40).launch()
+demo.queue(default_concurrency_limit=40).launch()

main_backend.py

@@ -0,0 +1,127 @@
+import argparse
+import logging
+import pprint
+import os
+
+from huggingface_hub import snapshot_download
+
+import src.backend.run_eval_suite as run_eval_suite
+import src.backend.manage_requests as manage_requests
+import src.backend.sort_queue as sort_queue
+import src.envs as envs
+
+os.environ['PYTORCH_CUDA_ALLOC_CONF'] = 'expandable_segments:True'
+
+logging.basicConfig(level=logging.ERROR)
+pp = pprint.PrettyPrinter(width=80)
+
+PENDING_STATUS = "PENDING"
+RUNNING_STATUS = "RUNNING"
+FINISHED_STATUS = "FINISHED"
+FAILED_STATUS = "FAILED"
+# import os
+# os.environ["HF_ENDPOINT"] = "https://hf-mirror.com"
+# snapshot_download(repo_id=envs.RESULTS_REPO, revision="main",
+#                 local_dir=envs.EVAL_RESULTS_PATH_BACKEND, repo_type="dataset", max_workers=60)
+
+# snapshot_download(repo_id=envs.QUEUE_REPO, revision="main",
+#                 local_dir=envs.EVAL_REQUESTS_PATH_BACKEND, repo_type="dataset", max_workers=60)
+# exit()
+
+def run_auto_eval(args):
+    if not args.reproduce:
+        current_pending_status = [PENDING_STATUS]
+        print('_________________')
+        manage_requests.check_completed_evals(
+            api=envs.API,
+            checked_status=RUNNING_STATUS,
+            completed_status=FINISHED_STATUS,
+            failed_status=FAILED_STATUS,
+            hf_repo=envs.QUEUE_REPO,
+            local_dir=envs.EVAL_REQUESTS_PATH_BACKEND,
+            hf_repo_results=envs.RESULTS_REPO,
+            local_dir_results=envs.EVAL_RESULTS_PATH_BACKEND
+        )
+        logging.info("Checked completed evals")
+        eval_requests = manage_requests.get_eval_requests(job_status=current_pending_status,
+                                                        hf_repo=envs.QUEUE_REPO,
+                                                        local_dir=envs.EVAL_REQUESTS_PATH_BACKEND)
+        logging.info("Got eval requests")
+        eval_requests = sort_queue.sort_models_by_priority(api=envs.API, models=eval_requests)
+        logging.info("Sorted eval requests")
+
+        print(f"Found {len(eval_requests)} {','.join(current_pending_status)} eval requests")
+        print(eval_requests)
+        if len(eval_requests) == 0:
+            print("No eval requests found. Exiting.")
+            return
+
+        if args.model is not None:
+            eval_request = manage_requests.EvalRequest(
+                model=args.model,
+                status=PENDING_STATUS,
+                precision=args.precision
+            )
+            pp.pprint(eval_request)
+        else:
+            eval_request = eval_requests[0]
+            pp.pprint(eval_request)
+
+        # manage_requests.set_eval_request(
+        #     api=envs.API,
+        #     eval_request=eval_request,
+        #     new_status=RUNNING_STATUS,
+        #     hf_repo=envs.QUEUE_REPO,
+        #     local_dir=envs.EVAL_REQUESTS_PATH_BACKEND
+        # )
+        # logging.info("Set eval request to running, now running eval")
+
+        run_eval_suite.run_evaluation(
+            eval_request=eval_request,
+            local_dir=envs.EVAL_RESULTS_PATH_BACKEND,
+            results_repo=envs.RESULTS_REPO,
+            batch_size=1,
+            device=envs.DEVICE,
+            no_cache=True,
+            need_check=not args.publish,
+            write_results=args.update
+        )
+        logging.info("Eval finished, now setting status to finished")
+    else:
+        eval_request = manage_requests.EvalRequest(
+            model=args.model,
+            status=PENDING_STATUS,
+            precision=args.precision
+        )
+        pp.pprint(eval_request)
+        logging.info("Running reproducibility eval")
+
+        run_eval_suite.run_evaluation(
+            eval_request=eval_request,
+            local_dir=envs.EVAL_RESULTS_PATH_BACKEND,
+            results_repo=envs.RESULTS_REPO,
+            batch_size=1,
+            device=envs.DEVICE,
+            need_check=not args.publish,
+            write_results=args.update
+        )
+        logging.info("Reproducibility eval finished")
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Run auto evaluation with optional reproducibility feature")
+
+    # Optional arguments
+    parser.add_argument("--reproduce", type=bool, default=True, help="Reproduce the evaluation results")
+    parser.add_argument("--model", type=str, default=None, help="Your Model ID")
+    parser.add_argument("--precision", type=str, default="float16", help="Precision of your model")
+    parser.add_argument("--publish", type=bool, default=False, help="whether directly publish the evaluation results on HF")
+    parser.add_argument("--update", type=bool, default=False, help="whether to update google drive files")
+
+    args = parser.parse_args()
+
+    run_auto_eval(args)
+
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -1,16 +1,17 @@
-APScheduler
-black
-datasets
-gradio
-gradio[oauth]
-gradio_leaderboard==0.0.9
-gradio_client
+APScheduler==3.10.1
+black==23.11.0
+click==8.1.3
+datasets==2.14.5
+gradio==4.4.0
+gradio_client==0.7.0
 huggingface-hub>=0.18.0
-matplotlib
-numpy
-pandas
-python-dateutil
-tqdm
-transformers
+litellm==1.15.1
+matplotlib==3.7.1
+numpy==1.24.2
+pandas==2.0.0
+python-dateutil==2.8.2
+requests==2.28.2
+tqdm==4.65.0
+transformers==4.35.2
 tokenizers>=0.15.0
-sentencepiece
+sentence-transformers==2.2.2

scripts/create_request_file.py

@@ -0,0 +1,110 @@
+import json
+import os
+import pprint
+import re
+from datetime import datetime, timezone
+
+import click
+from colorama import Fore
+from huggingface_hub import HfApi, snapshot_download
+
+from src.envs import QUEUE_REPO, EVAL_REQUESTS_PATH
+
+
+precisions = ("float16", "bfloat16", "8bit (LLM.int8)", "4bit (QLoRA / FP4)", "GPTQ")
+model_types = ("pretrained", "fine-tuned", "RL-tuned", "instruction-tuned")
+weight_types = ("Original", "Delta", "Adapter")
+
+
+def get_model_size(model_info, precision: str):
+    size_pattern = size_pattern = re.compile(r"(\d\.)?\d+(b|m)")
+    try:
+        model_size = round(model_info.safetensors["total"] / 1e9, 3)
+    except (AttributeError, TypeError):
+        try:
+            size_match = re.search(size_pattern, model_info.modelId.lower())
+            model_size = size_match.group(0)
+            model_size = round(float(model_size[:-1]) if model_size[-1] == "b"
+                            else float(model_size[:-1]) / 1e3, 3)
+        except AttributeError:
+            return 0  # Unknown model sizes are indicated as 0, see NUMERIC_INTERVALS in app.py
+
+    size_factor = 8 if (precision == "GPTQ" or "gptq" in model_info.modelId.lower()) else 1
+    model_size = size_factor * model_size
+    return model_size
+
+
+def main():
+    api = HfApi()
+    current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
+    snapshot_download(repo_id=QUEUE_REPO, revision="main", local_dir=EVAL_REQUESTS_PATH,
+                    repo_type="dataset")
+
+    model_name = click.prompt("Enter model name")
+    revision = click.prompt("Enter revision", default="main")
+    precision = click.prompt("Enter precision", default="float16", type=click.Choice(precisions))
+    model_type = click.prompt("Enter model type", type=click.Choice(model_types))
+    weight_type = click.prompt("Enter weight type", default="Original",
+                            type=click.Choice(weight_types))
+    base_model = click.prompt("Enter base model", default="")
+    status = click.prompt("Enter status", default="FINISHED")
+
+    try:
+        model_info = api.model_info(repo_id=model_name, revision=revision)
+    except Exception as e:
+        print(f"{Fore.RED}Could not find model info for {model_name} on the Hub\n{e}{Fore.RESET}")
+        return 1
+
+    model_size = get_model_size(model_info=model_info, precision=precision)
+
+    try:
+        license = model_info.cardData["license"]
+    except Exception:
+        license = "?"
+
+    eval_entry = {
+        "model": model_name,
+        "base_model": base_model,
+        "revision": revision,
+        "private": False,
+        "precision": precision,
+        "weight_type": weight_type,
+        "status": status,
+        "submitted_time": current_time,
+        "model_type": model_type,
+        "likes": model_info.likes,
+        "params": model_size,
+        "license": license,
+    }
+
+    user_name = ""
+    model_path = model_name
+    if "/" in model_name:
+        user_name = model_name.split("/")[0]
+        model_path = model_name.split("/")[1]
+
+    pprint.pprint(eval_entry)
+
+    if click.confirm("Do you want to continue? This request file will be pushed to the hub"):
+        click.echo("continuing...")
+
+        out_dir = f"{EVAL_REQUESTS_PATH}/{user_name}"
+        os.makedirs(out_dir, exist_ok=True)
+        out_path = f"{out_dir}/{model_path}_eval_request_{False}_{precision}_{weight_type}.json"
+
+        with open(out_path, "w") as f:
+            f.write(json.dumps(eval_entry))
+
+        api.upload_file(
+            path_or_fileobj=out_path,
+            path_in_repo=out_path.split(f"{EVAL_REQUESTS_PATH}/")[1],
+            repo_id=QUEUE_REPO,
+            repo_type="dataset",
+            commit_message=f"Add {model_name} to eval queue",
+        )
+    else:
+        click.echo("aborting...")
+
+
+if __name__ == "__main__":
+    main()

tests/test_evaluate_model.py

@@ -0,0 +1,87 @@
+import unittest
+from unittest.mock import patch
+
+import pandas as pd
+
+import src.backend.evaluate_model as evaluate_model
+import src.envs as envs
+
+
+class TestEvaluator(unittest.TestCase):
+
+    def setUp(self):
+        self.model_name = 'test_model'
+        self.revision = 'test_revision'
+        self.precision = 'test_precision'
+        self.batch_size = 10
+        self.device = 'test_device'
+        self.no_cache = False
+        self.limit = 10
+
+    @patch('src.backend.evaluate_model.SummaryGenerator')
+    @patch('src.backend.evaluate_model.EvaluationModel')
+    def test_evaluator_initialization(self, mock_eval_model, mock_summary_generator):
+        evaluator = evaluate_model.Evaluator(self.model_name, self.revision,
+                                            self.precision, self.batch_size,
+                                            self.device, self.no_cache, self.limit)
+
+        mock_summary_generator.assert_called_once_with(self.model_name, self.revision)
+        mock_eval_model.assert_called_once_with(envs.HEM_PATH)
+        self.assertEqual(evaluator.model, self.model_name)
+
+    @patch('src.backend.evaluate_model.EvaluationModel')
+    @patch('src.backend.evaluate_model.SummaryGenerator')
+    def test_evaluator_initialization_error(self, mock_summary_generator, mock_eval_model):
+        mock_eval_model.side_effect = Exception('test_exception')
+        with self.assertRaises(Exception):
+            evaluate_model.Evaluator(self.model_name, self.revision,
+                                    self.precision, self.batch_size,
+                                    self.device, self.no_cache, self.limit)
+
+    @patch('src.backend.evaluate_model.SummaryGenerator')
+    @patch('src.backend.evaluate_model.EvaluationModel')
+    @patch('src.backend.evaluate_model.pd.read_csv')
+    @patch('src.backend.util.format_results')
+    def test_evaluate_method(self, mock_format_results, mock_read_csv, mock_eval_model,
+                            mock_summary_generator):
+        evaluator = evaluate_model.Evaluator(self.model_name, self.revision,
+                                            self.precision, self.batch_size,
+                                            self.device, self.no_cache, self.limit)
+
+        # Mock setup
+        mock_format_results.return_value = {'test': 'result'}
+        mock_read_csv.return_value = pd.DataFrame({'column1': ['data1', 'data2']})
+        mock_summary_generator.return_value.generate_summaries.return_value = pd.DataFrame({'column1': ['summary1', 'summary2']})
+        mock_summary_generator.return_value.avg_length = 100
+        mock_summary_generator.return_value.answer_rate = 1.0
+        mock_summary_generator.return_value.error_rate = 0.0
+        mock_eval_model.return_value.compute_accuracy.return_value = 1.0
+        mock_eval_model.return_value.hallucination_rate = 0.0
+        mock_eval_model.return_value.evaluate_hallucination.return_value = [0.5]
+
+        # Method call and assertions
+        results = evaluator.evaluate()
+        mock_format_results.assert_called_once_with(model_name=self.model_name,
+                                                    revision=self.revision,
+                                                    precision=self.precision,
+                                                    accuracy=1.0, hallucination_rate=0.0,
+                                                    answer_rate=1.0, avg_summary_len=100,
+                                                    error_rate=0.0)
+        mock_read_csv.assert_called_once_with(envs.SOURCE_PATH)
+
+    @patch('src.backend.evaluate_model.SummaryGenerator')
+    @patch('src.backend.evaluate_model.EvaluationModel')
+    @patch('src.backend.evaluate_model.pd.read_csv')
+    def test_evaluate_with_file_not_found(self, mock_read_csv, mock_eval_model,
+                                        mock_summary_generator):
+        mock_read_csv.side_effect = FileNotFoundError('test_exception')
+        evaluator = evaluate_model.Evaluator(self.model_name, self.revision,
+                                            self.precision, self.batch_size,
+                                            self.device, self.no_cache, self.limit)
+
+        with self.assertRaises(FileNotFoundError):
+            evaluator.evaluate()
+
+
+if __name__ == '__main__':
+    unittest.main()

tests/test_evaluator.py

@@ -0,0 +1,59 @@
+import unittest
+from unittest.mock import patch
+
+import pandas as pd
+
+import src.backend.model_operations as model_operations
+
+
+class TestEvaluator(unittest.TestCase):
+
+    def setUp(self):
+        self.model_path = "test_model"
+
+    @patch("src.backend.model_operations.load_evaluation_model")
+    def test_init(self, mock_load_evaluation_model):
+        model_operations.EvaluationModel(self.model_path)
+        mock_load_evaluation_model.assert_called_once_with(self.model_path)
+
+    @patch("src.backend.model_operations.load_evaluation_model")
+    def test_evaluate_hallucination(self, mock_load_evaluation_model):
+        model = model_operations.EvaluationModel(self.model_path)
+        df = pd.DataFrame({'source': ['source1', 'source2'], 'summary': ['summary1', 'summary2']})
+
+        mock_load_evaluation_model.return_value.predict.return_value = [0.8, 0.2]
+
+        scores = model.evaluate_hallucination(df)
+        self.assertEqual(scores, [0.8, 0.2])
+
+    @patch("src.backend.model_operations.load_evaluation_model")
+    def test_evaluate_hallucination_exception(self, mock_load_evaluation_model):
+        model = model_operations.EvaluationModel(self.model_path)
+        df = pd.DataFrame({'source': ['source1', 'source2'], 'summary': ['summary1', 'summary2']})
+
+        mock_load_evaluation_model.return_value.predict.side_effect = Exception("Test exception")
+
+        with self.assertRaises(Exception):
+            scores = model.evaluate_hallucination(df)
+
+    @patch("src.backend.model_operations.load_evaluation_model")
+    def test_compute_accuracy(self, mock_load_evaluation_model):
+        model = model_operations.EvaluationModel(self.model_path)
+        model.scores = [0.8, 0.2]
+
+        accuracy = model.compute_accuracy()
+        expected_accuracy = 50.0
+        self.assertEqual(accuracy, expected_accuracy)
+
+
+class TestLoadEvaluationModel(unittest.TestCase):
+
+    @patch("src.backend.model_operations.CrossEncoder")
+    def test_load_evaluation_model(self, mock_cross_encoder):
+        model_path = 'test_model_path'
+        model_operations.load_evaluation_model(model_path)
+        mock_cross_encoder.assert_called_once_with(model_path)
+
+
+if __name__ == '__main__':
+    unittest.main()

tests/test_main_backend.py

@@ -0,0 +1,54 @@
+import unittest
+from unittest.mock import patch
+
+import main_backend
+import src.backend.manage_requests as manage_requests
+
+
+class TestMainBackend(unittest.TestCase):
+
+    @patch('src.backend.manage_requests.check_completed_evals')
+    @patch('src.backend.manage_requests.get_eval_requests')
+    @patch('src.backend.sort_queue.sort_models_by_priority')
+    @patch('src.backend.manage_requests.set_eval_request')
+    @patch('src.backend.run_eval_suite.run_evaluation')
+    def test_run_auto_eval_with_pending_requests(self, mock_run_evaluation, mock_set_eval_request,
+                            mock_sort_models_by_priority, mock_get_eval_requests,
+                            mock_check_completed_evals):
+        mock_sort_models_by_priority.return_value = [manage_requests.EvalRequest(
+            model="test_model",
+            private=True,
+            status="PENDING",
+            json_filepath="test_filepath",
+            weight_type="test_weight_type",
+            precision="test_precision",
+            base_model="test_base_model",
+            revision="test_revision",
+        )]
+
+        main_backend.run_auto_eval()
+
+        # Assertions
+        mock_check_completed_evals.assert_called()
+        mock_get_eval_requests.assert_called()
+        mock_sort_models_by_priority.assert_called()
+        mock_set_eval_request.assert_called()
+        mock_run_evaluation.assert_called()
+
+    @patch('builtins.print')
+    @patch('src.backend.manage_requests.check_completed_evals')
+    @patch('src.backend.manage_requests.get_eval_requests')
+    def test_run_auto_eval_with_no_pending_requests(self, mock_get_eval_requests,
+                                                    mock_check_completed_evals, mock_print):
+        mock_get_eval_requests.return_value = []
+
+        main_backend.run_auto_eval()
+
+        # Assertions
+        mock_check_completed_evals.assert_called()
+        mock_get_eval_requests.assert_called()
+        mock_print.assert_any_call("No eval requests found. Exiting.")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_summary_generator.py

@@ -0,0 +1,68 @@
+import unittest
+from unittest.mock import patch
+
+import pandas as pd
+
+import src.backend.evaluate_model as evaluate_model
+
+
+class TestSummaryGenerator(unittest.TestCase):
+
+    def setUp(self):
+        self.model_id = "test_model"
+        self.revision = "test_revision"
+
+    @patch("src.backend.model_operations.AutoTokenizer")
+    @patch("src.backend.model_operations.AutoModelForCausalLM")
+    def test_init(self, mock_model, mock_tokenizer):
+        evaluate_model.SummaryGenerator(self.model_id, self.revision)
+        mock_tokenizer.from_pretrained.assert_called_once_with(self.model_id,
+                                                            self.revision)
+        mock_model.from_pretrained.assert_called_once_with(self.model_id,
+                                                        self.revision)
+
+    @patch("src.backend.model_operations.nlp")
+    @patch("src.backend.model_operations.AutoTokenizer")
+    @patch("src.backend.model_operations.AutoModelForCausalLM")
+    def test_generate_summaries(self, mock_model, mock_tokenizer, mock_nlp):
+        df = pd.DataFrame({'text': ['text1', 'text2'],
+                        'dataset': ['dataset1', 'dataset2']})
+
+        generator = evaluate_model.SummaryGenerator(self.model_id, self.revision)
+        generator.generate_summaries(df)
+
+        self.assertEqual(len(generator.summaries_df), len(df))
+
+    @patch("src.backend.model_operations.AutoTokenizer")
+    @patch("src.backend.model_operations.AutoModelForCausalLM")
+    def test_compute_avg_length(self, mock_model, mock_tokenizer):
+        generator = evaluate_model.SummaryGenerator(self.model_id, self.revision)
+        test_df = pd.DataFrame({'source': ['text'], 'summary': ['This is a test.'],
+                                'dataset': ['dataset']})
+        generator.summaries_df = test_df
+        generator._compute_avg_length()
+        self.assertEqual(generator.avg_length, 4)
+
+    @patch("src.backend.model_operations.AutoTokenizer")
+    @patch("src.backend.model_operations.AutoModelForCausalLM")
+    def test_compute_answer_rate(self, mock_model, mock_tokenizer):
+        generator = evaluate_model.SummaryGenerator(self.model_id, self.revision)
+        test_df = pd.DataFrame({'source': ['text'], 'summary': ['This is a test.'],
+                                'dataset': ['dataset']})
+        generator.summaries_df = test_df
+        generator._compute_answer_rate()
+        self.assertEqual(generator.answer_rate, 1)
+
+    @patch("src.backend.model_operations.AutoTokenizer")
+    @patch("src.backend.model_operations.AutoModelForCausalLM")
+    def test_error_rate(self, mock_model, mock_tokenizer):
+        generator = evaluate_model.SummaryGenerator(self.model_id, self.revision)
+        test_df = pd.DataFrame({'source': ['text'], 'summary': ['This is a test.'],
+                                'dataset': ['dataset']})
+        generator.summaries_df = test_df
+        generator._compute_error_rate(0)
+        self.assertEqual(generator.error_rate, 0)
+
+
+if __name__ == "__main__":
+    unittest.main()