evalscope 0.16.0__tar.gz → 0.16.2__tar.gz

{evalscope-0.16.0/evalscope.egg-info → evalscope-0.16.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: evalscope
-Version: 0.16.0
+Version: 0.16.2
 Summary: EvalScope: Lightweight LLMs Evaluation Framework
 Home-page: https://github.com/modelscope/evalscope
 Author: ModelScope team
@@ -17,12 +17,12 @@ Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: accelerate
-Requires-Dist: datasets<=3.2.0,>=3.0.0
+Requires-Dist: datasets>=3.0
 Requires-Dist: immutabledict
 Requires-Dist: jieba
 Requires-Dist: jsonlines
 Requires-Dist: langdetect
-Requires-Dist: latex2sympy2
+Requires-Dist: latex2sympy2_extended
 Requires-Dist: matplotlib
 Requires-Dist: modelscope[framework]
 Requires-Dist: nltk>=3.9
@@ -40,20 +40,19 @@ Requires-Dist: seaborn
 Requires-Dist: sympy
 Requires-Dist: tabulate
 Requires-Dist: torch
-Requires-Dist: torchvision
 Requires-Dist: tqdm
 Requires-Dist: transformers>=4.33
 Requires-Dist: word2number
 Provides-Extra: opencompass
-Requires-Dist: ms-opencompass>=0.1.4; extra == "opencompass"
+Requires-Dist: ms-opencompass>=0.1.6; extra == "opencompass"
 Provides-Extra: vlmeval
-Requires-Dist: ms-vlmeval>=0.0.9; extra == "vlmeval"
+Requires-Dist: ms-vlmeval>=0.0.17; extra == "vlmeval"
 Provides-Extra: rag
 Requires-Dist: langchain<0.4.0,>=0.3.0; extra == "rag"
 Requires-Dist: langchain-community<0.4.0,>=0.3.0; extra == "rag"
 Requires-Dist: langchain-core<0.4.0,>=0.3.0; extra == "rag"
 Requires-Dist: langchain-openai<0.4.0,>=0.3.0; extra == "rag"
-Requires-Dist: mteb==1.19.4; extra == "rag"
+Requires-Dist: mteb==1.38.20; extra == "rag"
 Requires-Dist: ragas==0.2.14; extra == "rag"
 Requires-Dist: webdataset>0.2.0; extra == "rag"
 Provides-Extra: perf
@@ -73,14 +72,15 @@ Requires-Dist: iopath; extra == "aigc"
 Requires-Dist: omegaconf; extra == "aigc"
 Requires-Dist: open_clip_torch; extra == "aigc"
 Requires-Dist: opencv-python; extra == "aigc"
+Requires-Dist: torchvision; extra == "aigc"
 Provides-Extra: all
 Requires-Dist: accelerate; extra == "all"
-Requires-Dist: datasets<=3.2.0,>=3.0.0; extra == "all"
+Requires-Dist: datasets>=3.0; extra == "all"
 Requires-Dist: immutabledict; extra == "all"
 Requires-Dist: jieba; extra == "all"
 Requires-Dist: jsonlines; extra == "all"
 Requires-Dist: langdetect; extra == "all"
-Requires-Dist: latex2sympy2; extra == "all"
+Requires-Dist: latex2sympy2_extended; extra == "all"
 Requires-Dist: matplotlib; extra == "all"
 Requires-Dist: modelscope[framework]; extra == "all"
 Requires-Dist: nltk>=3.9; extra == "all"
@@ -98,17 +98,16 @@ Requires-Dist: seaborn; extra == "all"
 Requires-Dist: sympy; extra == "all"
 Requires-Dist: tabulate; extra == "all"
 Requires-Dist: torch; extra == "all"
-Requires-Dist: torchvision; extra == "all"
 Requires-Dist: tqdm; extra == "all"
 Requires-Dist: transformers>=4.33; extra == "all"
 Requires-Dist: word2number; extra == "all"
-Requires-Dist: ms-opencompass>=0.1.4; extra == "all"
-Requires-Dist: ms-vlmeval>=0.0.9; extra == "all"
+Requires-Dist: ms-opencompass>=0.1.6; extra == "all"
+Requires-Dist: ms-vlmeval>=0.0.17; extra == "all"
 Requires-Dist: langchain<0.4.0,>=0.3.0; extra == "all"
 Requires-Dist: langchain-community<0.4.0,>=0.3.0; extra == "all"
 Requires-Dist: langchain-core<0.4.0,>=0.3.0; extra == "all"
 Requires-Dist: langchain-openai<0.4.0,>=0.3.0; extra == "all"
-Requires-Dist: mteb==1.19.4; extra == "all"
+Requires-Dist: mteb==1.38.20; extra == "all"
 Requires-Dist: ragas==0.2.14; extra == "all"
 Requires-Dist: webdataset>0.2.0; extra == "all"
 Requires-Dist: aiohttp; extra == "all"
@@ -125,6 +124,7 @@ Requires-Dist: iopath; extra == "all"
 Requires-Dist: omegaconf; extra == "all"
 Requires-Dist: open_clip_torch; extra == "all"
 Requires-Dist: opencv-python; extra == "all"
+Requires-Dist: torchvision; extra == "all"
 
 <p align="center">
     <br>
@@ -230,6 +230,9 @@ Please scan the QR code below to join our community groups:
 
 ## 🎉 News
 
+- 🔥 **[2025.06.19]** Added support for the BFCL-v3 benchmark, designed to evaluate model function-calling capabilities across various scenarios. For more information, refer to the [documentation](https://evalscope.readthedocs.io/zh-cn/latest/third_party/bfcl_v3.html).
+- 🔥 **[2025.06.02]** Added support for the Needle-in-a-Haystack test. Simply specify `needle_haystack` to conduct the test, and a corresponding heatmap will be generated in the `outputs/reports` folder, providing a visual representation of the model's performance. Refer to the [documentation](https://evalscope.readthedocs.io/en/latest/third_party/needle_haystack.html) for more details.
+- 🔥 **[2025.05.29]** Added support for two long document evaluation benchmarks: [DocMath](https://modelscope.cn/datasets/yale-nlp/DocMath-Eval/summary) and [FRAMES](https://modelscope.cn/datasets/iic/frames/summary). For usage guidelines, please refer to the [documentation](https://evalscope.readthedocs.io/en/latest/get_started/supported_dataset.html).
 - 🔥 **[2025.05.16]** Model service performance stress testing now supports setting various levels of concurrency and outputs a performance test report. [Reference example](https://evalscope.readthedocs.io/en/latest/user_guides/stress_test/quick_start.html#id3).
 - 🔥 **[2025.05.13]** Added support for the [ToolBench-Static](https://modelscope.cn/datasets/AI-ModelScope/ToolBench-Static) dataset to evaluate model's tool-calling capabilities. Refer to the [documentation](https://evalscope.readthedocs.io/en/latest/third_party/toolbench.html) for usage instructions. Also added support for the [DROP](https://modelscope.cn/datasets/AI-ModelScope/DROP/dataPeview) and [Winogrande](https://modelscope.cn/datasets/AI-ModelScope/winogrande_val) benchmarks to assess the reasoning capabilities of models.
 - 🔥 **[2025.04.29]** Added Qwen3 Evaluation Best Practices, [welcome to read 📖](https://evalscope.readthedocs.io/en/latest/best_practice/qwen3.html)

{evalscope-0.16.0 → evalscope-0.16.2}/README.md

@@ -102,6 +102,9 @@ Please scan the QR code below to join our community groups:
 
 ## 🎉 News
 
+- 🔥 **[2025.06.19]** Added support for the BFCL-v3 benchmark, designed to evaluate model function-calling capabilities across various scenarios. For more information, refer to the [documentation](https://evalscope.readthedocs.io/zh-cn/latest/third_party/bfcl_v3.html).
+- 🔥 **[2025.06.02]** Added support for the Needle-in-a-Haystack test. Simply specify `needle_haystack` to conduct the test, and a corresponding heatmap will be generated in the `outputs/reports` folder, providing a visual representation of the model's performance. Refer to the [documentation](https://evalscope.readthedocs.io/en/latest/third_party/needle_haystack.html) for more details.
+- 🔥 **[2025.05.29]** Added support for two long document evaluation benchmarks: [DocMath](https://modelscope.cn/datasets/yale-nlp/DocMath-Eval/summary) and [FRAMES](https://modelscope.cn/datasets/iic/frames/summary). For usage guidelines, please refer to the [documentation](https://evalscope.readthedocs.io/en/latest/get_started/supported_dataset.html).
 - 🔥 **[2025.05.16]** Model service performance stress testing now supports setting various levels of concurrency and outputs a performance test report. [Reference example](https://evalscope.readthedocs.io/en/latest/user_guides/stress_test/quick_start.html#id3).
 - 🔥 **[2025.05.13]** Added support for the [ToolBench-Static](https://modelscope.cn/datasets/AI-ModelScope/ToolBench-Static) dataset to evaluate model's tool-calling capabilities. Refer to the [documentation](https://evalscope.readthedocs.io/en/latest/third_party/toolbench.html) for usage instructions. Also added support for the [DROP](https://modelscope.cn/datasets/AI-ModelScope/DROP/dataPeview) and [Winogrande](https://modelscope.cn/datasets/AI-ModelScope/winogrande_val) benchmarks to assess the reasoning capabilities of models.
 - 🔥 **[2025.04.29]** Added Qwen3 Evaluation Best Practices, [welcome to read 📖](https://evalscope.readthedocs.io/en/latest/best_practice/qwen3.html)

evalscope-0.16.2/evalscope/app/__init__.py

@@ -0,0 +1,28 @@
+# Copyright (c) Alibaba, Inc. and its affiliates.
+from typing import TYPE_CHECKING
+
+from evalscope.utils.import_utils import _LazyModule
+
+if TYPE_CHECKING:
+    from .app import create_app
+    from .arguments import add_argument
+
+else:
+    _import_structure = {
+        'app': [
+            'create_app',
+        ],
+        'arguments': [
+            'add_argument',
+        ],
+    }
+
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()['__file__'],
+        _import_structure,
+        module_spec=__spec__,
+        extra_objects={},
+    )

{evalscope-0.16.0/evalscope/report → evalscope-0.16.2/evalscope/app}/app.py

@@ -1,6 +1,7 @@
 import argparse
 import glob
 import gradio as gr
+import json
 import numpy as np
 import os
 import pandas as pd
@@ -11,35 +12,15 @@ from dataclasses import dataclass
 from typing import Any, List, Union
 
 from evalscope.constants import DataCollection
-from evalscope.report import Report, ReportKey, add_argument, get_data_frame, get_report_list
+from evalscope.report import Report, ReportKey, get_data_frame, get_report_list
 from evalscope.utils.io_utils import OutputsStructure, yaml_to_dict
 from evalscope.utils.logger import configure_logging, get_logger
 from evalscope.version import __version__
+from .arguments import add_argument
+from .constants import DATASET_TOKEN, LATEX_DELIMITERS, MODEL_TOKEN, PLOTLY_THEME, REPORT_TOKEN
 
 logger = get_logger()
 
-PLOTLY_THEME = 'plotly_dark'
-REPORT_TOKEN = '@@'
-MODEL_TOKEN = '::'
-DATASET_TOKEN = ', '
-LATEX_DELIMITERS = [{
-    'left': '$$',
-    'right': '$$',
-    'display': True
-}, {
-    'left': '$',
-    'right': '$',
-    'display': False
-}, {
-    'left': '\\(',
-    'right': '\\)',
-    'display': False
-}, {
-    'left': '\\[',
-    'right': '\\]',
-    'display': True
-}]
-
 
 def scan_for_report_folders(root_path):
     """Scan for folders containing reports subdirectories"""
@@ -155,11 +136,11 @@ def plot_single_report_scores(df: pd.DataFrame):
 
 def plot_single_report_sunburst(report_list: List[Report]):
     if report_list[0].name == DataCollection.NAME:
-        df = get_data_frame(report_list)
+        df = get_data_frame(report_list=report_list)
         categories = sorted([i for i in df.columns if i.startswith(ReportKey.category_prefix)])
         path = categories + [ReportKey.subset_name]
     else:
-        df = get_data_frame(report_list, flatten_metrics=False)
+        df = get_data_frame(report_list=report_list, flatten_metrics=False)
         categories = sorted([i for i in df.columns if i.startswith(ReportKey.category_prefix)])
         path = [ReportKey.dataset_name] + categories + [ReportKey.subset_name]
     logger.debug(f'df: {df}')
@@ -185,6 +166,13 @@ def get_single_dataset_df(df: pd.DataFrame, dataset_name: str):
     return df, styler
 
 
+def get_report_analysis(report_list: List[Report], dataset_name: str) -> str:
+    for report in report_list:
+        if report.dataset_name == dataset_name:
+            return report.analysis
+    return 'N/A'
+
+
 def plot_single_dataset_scores(df: pd.DataFrame):
     # TODO: add metric radio and relace category name
     plot = px.bar(
@@ -246,7 +234,7 @@ def convert_html_tags(text):
 def process_string(string: str, max_length: int = 2048) -> str:
     string = convert_html_tags(string)  # for display labels e.g.
     if max_length and len(string) > max_length:
-        return f'{string[:max_length // 2]}......{string[-max_length // 2:]}'
+        return f'{string[:max_length // 2]}...[truncate]...{string[-max_length // 2:]}'
     return string
 
 
@@ -270,7 +258,7 @@ def dict_to_markdown(data) -> str:
     return '\n\n'.join(markdown_lines)
 
 
-def process_model_prediction(item: Any, max_length: int = 2048) -> str:
+def process_model_prediction_old(item: Any, max_length: int = 2048) -> str:
     """
     Process model prediction output into a formatted string.
 
@@ -294,6 +282,20 @@ def process_model_prediction(item: Any, max_length: int = 2048) -> str:
     return result
 
 
+def process_model_prediction(item: Any, max_length: int = 4096) -> str:
+    if isinstance(item, (dict, list)):
+        result = json.dumps(item, ensure_ascii=False, indent=2)
+        result = f'```json\n{result}\n```'
+    else:
+        result = str(item)
+
+    # Apply HTML tag conversion and truncation only at the final output
+    if max_length is not None:
+        return process_string(result, max_length)
+
+    return result
+
+
 def normalize_score(score):
     try:
         if isinstance(score, bool):
@@ -456,6 +458,10 @@ def create_single_model_tab(sidebar: SidebarComponents, lang: str):
             'zh': '数据集分数',
             'en': 'Dataset Scores'
         },
+        'report_analysis': {
+            'zh': '报告智能分析',
+            'en': 'Report Intelligent Analysis'
+        },
         'dataset_scores_table': {
             'zh': '数据集分数表',
             'en': 'Dataset Scores Table'
@@ -511,6 +517,9 @@ def create_single_model_tab(sidebar: SidebarComponents, lang: str):
     with gr.Tab(locale_dict['dataset_details'][lang]):
         dataset_radio = gr.Radio(
             label=locale_dict['select_dataset'][lang], choices=[], show_label=True, interactive=True)
+        # show dataset details
+        with gr.Accordion(locale_dict['report_analysis'][lang], open=True):
+            report_analysis = gr.Markdown(value='N/A', show_copy_button=True)
         gr.Markdown(f'### {locale_dict["dataset_scores"][lang]}')
         dataset_plot = gr.Plot(value=None, scale=1, label=locale_dict['dataset_scores'][lang])
         gr.Markdown(f'### {locale_dict["dataset_scores_table"][lang]}')
@@ -586,15 +595,16 @@ def create_single_model_tab(sidebar: SidebarComponents, lang: str):
     @gr.on(
         triggers=[dataset_radio.change, report_list.change],
         inputs=[dataset_radio, report_list],
-        outputs=[dataset_plot, dataset_table, subset_select, data_review_df])
+        outputs=[dataset_plot, dataset_table, subset_select, data_review_df, report_analysis])
     def update_single_report_dataset(dataset_name, report_list):
         logger.debug(f'Updating single report dataset: {dataset_name}')
-        report_df = get_data_frame(report_list)
+        report_df = get_data_frame(report_list=report_list)
+        analysis = get_report_analysis(report_list, dataset_name)
         data_score_df, styler = get_single_dataset_df(report_df, dataset_name)
         data_score_plot = plot_single_dataset_scores(data_score_df)
         subsets = data_score_df[ReportKey.subset_name].unique().tolist()
         logger.debug(f'subsets: {subsets}')
-        return data_score_plot, styler, gr.update(choices=subsets, value=None), None
+        return data_score_plot, styler, gr.update(choices=subsets, value=None), None, analysis
 
     @gr.on(
         triggers=[subset_select.change],

evalscope-0.16.2/evalscope/app/constants.py

@@ -0,0 +1,21 @@
+PLOTLY_THEME = 'plotly_dark'
+REPORT_TOKEN = '@@'
+MODEL_TOKEN = '::'
+DATASET_TOKEN = ', '
+LATEX_DELIMITERS = [{
+    'left': '$$',
+    'right': '$$',
+    'display': True
+}, {
+    'left': '$',
+    'right': '$',
+    'display': False
+}, {
+    'left': '\\(',
+    'right': '\\)',
+    'display': False
+}, {
+    'left': '\\[',
+    'right': '\\]',
+    'display': True
+}]

{evalscope-0.16.0 → evalscope-0.16.2}/evalscope/arguments.py

@@ -67,7 +67,7 @@ def add_argument(parser: argparse.ArgumentParser):
     parser.add_argument('--eval-config', type=str, required=False, help='The eval task config file path for evaluation backend.')  # noqa: E501
     parser.add_argument('--stage', type=str, default='all', help='The stage of evaluation pipeline.',
                         choices=[EvalStage.ALL, EvalStage.INFER, EvalStage.REVIEW])
-    parser.add_argument('--limit', type=int, default=None, help='Max evaluation samples num for each subset.')
+    parser.add_argument('--limit', type=float, default=None, help='Max evaluation samples num for each subset.')
     parser.add_argument('--eval-batch-size', type=int, default=1, help='The batch size for evaluation.')
 
     # Cache and working directory arguments
@@ -89,6 +89,7 @@ def add_argument(parser: argparse.ArgumentParser):
     parser.add_argument('--judge-strategy', type=str, default=JudgeStrategy.AUTO, help='The judge strategy.')
     parser.add_argument('--judge-model-args', type=json.loads, default='{}', help='The judge model args, should be a json string.')  # noqa: E501
     parser.add_argument('--judge-worker-num', type=int, default=1, help='The number of workers for the judge model.')
+    parser.add_argument('--analysis-report', action='store_true', default=False, help='Generate analysis report for the evaluation results using judge model.')  # noqa: E501
     # yapf: enable
 
 

{evalscope-0.16.0 → evalscope-0.16.2}/evalscope/backend/opencompass/backend_manager.py

@@ -1,4 +1,5 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
+import os
 import subprocess
 import tempfile
 from dataclasses import asdict
@@ -204,7 +205,7 @@ class OpenCompassBackendManager(BackendManager):
                     model_d['meta_template'] = get_template(model_d['meta_template'])
 
                 # set the 'abbr' as the 'path' if 'abbr' is not specified
-                model_d['abbr'] = model_d['path']
+                model_d['abbr'] = os.path.basename(model_d['path'])
 
                 model_config = ApiModelConfig(**model_d)
                 models.append(asdict(model_config))

{evalscope-0.16.0 → evalscope-0.16.2}/evalscope/backend/rag_eval/clip_benchmark/dataset_builder.py

@@ -1,4 +1,5 @@
 import os
+import posixpath  # For URL path handling
 import torch
 from torch.utils.data import DataLoader
 from torch.utils.data import Dataset as TorchDataset
@@ -186,42 +187,53 @@ def build_wds_dataset(dataset_name, transform, split='test', data_dir='root', ca
 
     Set `cache_dir` to a path to cache the dataset, otherwise, no caching will occur.
     """
+    import requests
     import webdataset as wds
 
     def read_txt(fname):
-        if '://' in fname:
-            stream = os.popen("curl -L -s --fail '%s'" % fname, 'r')
-            value = stream.read()
-            if stream.close():
-                raise FileNotFoundError('Failed to retreive data')
+        if fname.startswith(('http://', 'https://')):
+            try:
+                response = requests.get(fname)
+                response.raise_for_status()  # Ensure the HTTP request was successful
+                return response.text
+            except requests.exceptions.RequestException as e:
+                raise FileNotFoundError(f'Failed to read {fname}: {e}')
         else:
             with open(fname, 'r') as file:
-                value = file.read()
-        return value
+                return file.read()
+
+    def url_path_join(*parts):
+        """Join URL path parts with forward slashes regardless of platform"""
+        return posixpath.join(*parts)
 
     if not data_dir:
         data_dir = f'https://modelscope.cn/datasets/clip-benchmark/wds_{dataset_name}/resolve/master'
 
     # Git LFS files have a different file path to access the raw data than other files
-    if data_dir.startswith('https://modelscope.cn/datasets'):
+    is_url = data_dir.startswith(('http://', 'https://'))
+    if is_url and data_dir.startswith('https://modelscope.cn/datasets'):
         *split_url_head, _, url_path = data_dir.split('/', 7)
         url_head = '/'.join(split_url_head)
         metadata_dir = '/'.join([url_head, 'resolve', url_path])
         tardata_dir = '/'.join([url_head, 'resolve', url_path])
     else:
         metadata_dir = tardata_dir = data_dir
+
+    # Use appropriate path joining function based on whether we're dealing with a URL
+    path_join = url_path_join if is_url else os.path.join
+
     # Get number of shards
-    nshards_fname = os.path.join(metadata_dir, split, 'nshards.txt')
+    nshards_fname = path_join(metadata_dir, split, 'nshards.txt')
     nshards = int(read_txt(nshards_fname))  # Do not catch FileNotFound, nshards.txt should be mandatory
 
     # Get dataset type (classification or retrieval)
-    type_fname = os.path.join(metadata_dir, 'dataset_type.txt')
+    type_fname = path_join(metadata_dir, 'dataset_type.txt')
     try:
         dataset_type = read_txt(type_fname).strip().lower()
     except FileNotFoundError:
         dataset_type = 'classification'
 
-    filepattern = os.path.join(tardata_dir, split, '{0..%d}.tar' % (nshards - 1))
+    filepattern = path_join(tardata_dir, split, '{0..%d}.tar' % (nshards - 1))
     # Load webdataset (support WEBP, PNG, and JPG for now)
     if not cache_dir or not isinstance(cache_dir, str):
         cache_dir = None

{evalscope-0.16.0 → evalscope-0.16.2}/evalscope/backend/rag_eval/cmteb/arguments.py

@@ -11,7 +11,9 @@ class ModelArguments:
     pooling_mode: Optional[str] = None
     max_seq_length: int = 512  # max sequence length
     # prompt for llm based model
-    prompt: str = ''
+    prompt: Optional[str] = None
+    # prompts dictionary for different tasks, if prompt is not set
+    prompts: Optional[Dict[str, str]] = None
     # model kwargs
     model_kwargs: dict = field(default_factory=dict)
     # config kwargs
@@ -33,6 +35,7 @@ class ModelArguments:
             'pooling_mode': self.pooling_mode,
             'max_seq_length': self.max_seq_length,
             'prompt': self.prompt,
+            'prompts': self.prompts,
             'model_kwargs': self.model_kwargs,
             'config_kwargs': self.config_kwargs,
             'encode_kwargs': self.encode_kwargs,

{evalscope-0.16.0 → evalscope-0.16.2}/evalscope/backend/rag_eval/cmteb/task_template.py

@@ -1,6 +1,6 @@
 import mteb
 import os
-from mteb.task_selection import results_to_dataframe
+from tabulate import tabulate
 
 from evalscope.backend.rag_eval import EmbeddingModel, cmteb
 from evalscope.utils.logger import get_logger
@@ -12,14 +12,27 @@ def show_results(output_folder, model, results):
     model_name = model.mteb_model_meta.model_name_as_path()
     revision = model.mteb_model_meta.revision
 
-    results_df = results_to_dataframe({model_name: {revision: results}})
+    data = []
+    for model_res in results:
+        main_res = model_res.only_main_score()
+        for split, score in main_res.scores.items():
+            for sub_score in score:
+                data.append({
+                    'Model': model_name.replace('eval__', ''),
+                    'Revision': revision,
+                    'Task Type': main_res.task_type,
+                    'Task': main_res.task_name,
+                    'Split': split,
+                    'Subset': sub_score['hf_subset'],
+                    'Main Score': sub_score['main_score'],
+                })
 
     save_path = os.path.join(
         output_folder,
         model_name,
         revision,
     )
-    logger.info(f'Evaluation results:\n{results_df.to_markdown()}')
+    logger.info(f'Evaluation results:\n{tabulate(data, headers="keys", tablefmt="grid")}')
     logger.info(f'Evaluation results saved in {os.path.abspath(save_path)}')
 
 
@@ -34,6 +47,7 @@ def one_stage_eval(
     tasks = cmteb.TaskBase.get_tasks(task_names=eval_args['tasks'], dataset_path=custom_dataset_path)
     evaluation = mteb.MTEB(tasks=tasks)
 
+    eval_args['encode_kwargs'] = model_args.get('encode_kwargs', {})
     # run evaluation
     results = evaluation.run(model, **eval_args)
 
@@ -66,6 +80,7 @@ def two_stage_eval(
             overwrite_results=True,
             hub=eval_args['hub'],
             limits=eval_args['limits'],
+            encode_kwargs=model1_args.get('encode_kwargs', {}),
         )
         # stage 2: run cross encoder
         results = evaluation.run(
@@ -77,6 +92,7 @@ def two_stage_eval(
             overwrite_results=True,
             hub=eval_args['hub'],
             limits=eval_args['limits'],
+            encode_kwargs=model2_args.get('encode_kwargs', {}),
         )
 
         # save and log results