fusion-bench 0.2.19__py3-none-any.whl → 0.2.20__py3-none-any.whl

fusion_bench/modelpool/clip_vision/modelpool.py

@@ -1,6 +1,6 @@
 import logging
 from copy import deepcopy
-from typing import Optional, Union
+from typing import Literal, Optional, Union
 
 from datasets import load_dataset
 from lightning.fabric.utilities import rank_zero_only
@@ -11,6 +11,9 @@ from transformers import CLIPModel, CLIPProcessor, CLIPVisionModel
 from typing_extensions import override
 
 from fusion_bench.utils import instantiate, timeit_context
+from fusion_bench.utils.modelscope import (
+    resolve_repo_path,
+)
 
 from ..base_pool import BaseModelPool
 
@@ -25,25 +28,32 @@ class CLIPVisionModelPool(BaseModelPool):
     the specifics of the CLIP Vision models provided by the Hugging Face Transformers library.
     """
 
-    _config_mapping = BaseModelPool._config_mapping | {"_processor": "processor"}
+    _config_mapping = BaseModelPool._config_mapping | {
+        "_processor": "processor",
+        "_platform": "hf",
+    }
 
     def __init__(
         self,
         models: DictConfig,
         *,
         processor: Optional[DictConfig] = None,
+        platform: Literal["hf", "huggingface", "modelscope"] = "hf",
         **kwargs,
     ):
         super().__init__(models, **kwargs)
-
         self._processor = processor
+        self._platform = platform
 
     def load_processor(self, *args, **kwargs) -> CLIPProcessor:
         assert self._processor is not None, "Processor is not defined in the config"
         if isinstance(self._processor, str):
             if rank_zero_only.rank == 0:
                 log.info(f"Loading `transformers.CLIPProcessor`: {self._processor}")
-            processor = CLIPProcessor.from_pretrained(self._processor)
+            repo_path = resolve_repo_path(
+                repo_id=self._processor, repo_type="model", platform=self._platform
+            )
+            processor = CLIPProcessor.from_pretrained(repo_path, *args, **kwargs)
         else:
             processor = instantiate(self._processor, *args, **kwargs)
         return processor
@@ -54,7 +64,10 @@ class CLIPVisionModelPool(BaseModelPool):
         if isinstance(model_config, str):
             if rank_zero_only.rank == 0:
                 log.info(f"Loading `transformers.CLIPModel`: {model_config}")
-            clip_model = CLIPModel.from_pretrained(model_config, *args, **kwargs)
+            repo_path = resolve_repo_path(
+                repo_id=model_config, repo_type="model", platform=self._platform
+            )
+            clip_model = CLIPModel.from_pretrained(repo_path, *args, **kwargs)
             return clip_model
         else:
             assert isinstance(
@@ -107,14 +120,17 @@ class CLIPVisionModelPool(BaseModelPool):
             if isinstance(model, str):
                 if rank_zero_only.rank == 0:
                     log.info(f"Loading `transformers.CLIPVisionModel`: {model}")
-                return CLIPVisionModel.from_pretrained(model, *args, **kwargs)
+                repo_path = resolve_repo_path(
+                    model, repo_type="model", platform=self._platform
+                )
+                return CLIPVisionModel.from_pretrained(repo_path, *args, **kwargs)
             if isinstance(model, nn.Module):
                 if rank_zero_only.rank == 0:
                     log.info(f"Returning existing model: {model}")
                 return model
-
-        # If the model is not a string, we use the default load_model method
-        return super().load_model(model_name_or_config, *args, **kwargs)
+        else:
+            # If the model is not a string, we use the default load_model method
+            return super().load_model(model_name_or_config, *args, **kwargs)
 
     def load_train_dataset(self, dataset_name: str, *args, **kwargs):
         dataset_config = self._train_datasets[dataset_name]
@@ -123,7 +139,7 @@ class CLIPVisionModelPool(BaseModelPool):
                 log.info(
                     f"Loading train dataset using `datasets.load_dataset`: {dataset_config}"
                 )
-            dataset = load_dataset(dataset_config, split="train")
+            dataset = self._load_dataset(dataset_config, split="train")
         else:
             dataset = super().load_train_dataset(dataset_name, *args, **kwargs)
         return dataset
@@ -135,7 +151,7 @@ class CLIPVisionModelPool(BaseModelPool):
                 log.info(
                     f"Loading validation dataset using `datasets.load_dataset`: {dataset_config}"
                 )
-            dataset = load_dataset(dataset_config, split="validation")
+            dataset = self._load_dataset(dataset_config, split="validation")
         else:
             dataset = super().load_val_dataset(dataset_name, *args, **kwargs)
         return dataset
@@ -147,7 +163,24 @@ class CLIPVisionModelPool(BaseModelPool):
                 log.info(
                     f"Loading test dataset using `datasets.load_dataset`: {dataset_config}"
                 )
-            dataset = load_dataset(dataset_config, split="test")
+            dataset = self._load_dataset(dataset_config, split="test")
         else:
             dataset = super().load_test_dataset(dataset_name, *args, **kwargs)
         return dataset
+
+    def _load_dataset(self, name: str, split: str):
+        """
+        Load a dataset by its name and split.
+
+        Args:
+            dataset_name (str): The name of the dataset.
+            split (str): The split of the dataset to load (e.g., "train", "validation", "test").
+
+        Returns:
+            Dataset: The loaded dataset.
+        """
+        datset_dir = resolve_repo_path(
+            name, repo_type="dataset", platform=self._platform
+        )
+        dataset = load_dataset(datset_dir, split=split)
+        return dataset

fusion_bench/scripts/cli.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-This is the CLI script that is executed when the user runs the `fusion-bench` command.
+This is the CLI script that is executed when the user runs the `fusion_bench` command.
 The script is responsible for parsing the command-line arguments, loading the configuration file, and running the fusion algorithm.
 """
 

fusion_bench/utils/misc.py

@@ -1,6 +1,13 @@
-from typing import Iterable, List
+from difflib import get_close_matches
+from typing import Any, Iterable, List, Optional
 
-__all__ = ["first", "has_length", "join_list", "attr_equal"]
+__all__ = [
+    "first",
+    "has_length",
+    "join_list",
+    "attr_equal",
+    "validate_and_suggest_corrections",
+]
 
 
 def first(iterable: Iterable):
@@ -41,3 +48,42 @@ def attr_equal(obj, attr: str, value):
     if not hasattr(obj, attr):
         return False
     return getattr(obj, attr) == value
+
+
+def validate_and_suggest_corrections(
+    obj: Any, values: Iterable[Any], *, max_suggestions: int = 3, cutoff: float = 0.6
+) -> Any:
+    """
+    Return *obj* if it is contained in *values*.
+    Otherwise raise a helpful ``ValueError`` that lists the closest matches.
+
+    Args:
+        obj : Any
+            The value to validate.
+        values : Iterable[Any]
+            The set of allowed values.
+        max_suggestions : int, optional
+            How many typo-hints to include at most (default 3).
+        cutoff : float, optional
+            Similarity threshold for suggestions (0.0-1.0, default 0.6).
+
+    Returns:
+        The original *obj* if it is valid.
+
+    Raises:
+        ValueError: With a friendly message that points out possible typos.
+    """
+    # Normalise to a list so we can reuse it
+    value_list = list(values)
+
+    if obj in value_list:
+        return obj
+
+    # Build suggestions
+    str_values = list(map(str, value_list))
+    matches = get_close_matches(str(obj), str_values, n=max_suggestions, cutoff=cutoff)
+
+    msg = f"Invalid value {obj!r}. Allowed values: {value_list}"
+    if matches:
+        msg += f". Did you mean {', '.join(repr(m) for m in matches)}?"
+    raise ValueError(msg)

fusion_bench/utils/modelscope.py

@@ -0,0 +1,146 @@
+import os
+from typing import Literal, Optional
+
+from datasets import load_dataset as datasets_load_dataset
+from fusion_bench.utils import validate_and_suggest_corrections
+
+try:
+    from modelscope import snapshot_download as modelscope_snapshot_download
+except ImportError:
+
+    def modelscope_snapshot_download(*args, **kwargs):
+        raise ImportError(
+            "ModelScope is not installed. Please install it using `pip install modelscope` to use ModelScope models."
+        )
+
+
+try:
+    from huggingface_hub import snapshot_download as huggingface_snapshot_download
+except ImportError:
+
+    def huggingface_snapshot_download(*args, **kwargs):
+        raise ImportError(
+            "Hugging Face Hub is not installed. Please install it using `pip install huggingface_hub` to use Hugging Face models."
+        )
+
+
+__all__ = [
+    "load_dataset",
+    "resolve_repo_path",
+]
+
+AVAILABLE_PLATFORMS = ["hf", "huggingface", "modelscope"]
+
+
+def load_dataset(
+    name: str,
+    split: str = "train",
+    platform: Literal["hf", "huggingface", "modelscope"] = "hf",
+):
+    """
+    Load a dataset from Hugging Face or ModelScope.
+
+    Args:
+        platform (Literal['hf', 'modelscope']): The platform to load the dataset from.
+        name (str): The name of the dataset.
+        split (str): The split of the dataset to load (default is "train").
+
+    Returns:
+        Dataset: The loaded dataset.
+    """
+    validate_and_suggest_corrections(platform, AVAILABLE_PLATFORMS)
+    if platform == "hf" or platform == "huggingface":
+        return datasets_load_dataset(name, split=split)
+    elif platform == "modelscope":
+        dataset_dir = modelscope_snapshot_download(name, repo_type="dataset")
+        return datasets_load_dataset(dataset_dir, split=split)
+    else:
+        raise ValueError(
+            f"Unsupported platform: {platform}. Supported platforms are 'hf', 'huggingface', and 'modelscope'."
+        )
+
+
+def resolve_repo_path(
+    repo_id: str,
+    repo_type: Optional[str] = "model",
+    platform: Literal["hf", "huggingface", "modelscope"] = "hf",
+    **kwargs,
+):
+    """
+    Resolve and download a repository from various platforms to a local path.
+
+    This function handles multiple repository sources including local paths, Hugging Face,
+    and ModelScope. It automatically downloads remote repositories to local cache and
+    returns the local path for further use.
+
+    Args:
+        repo_id (str): Repository identifier. Can be:
+            - Local file/directory path (returned as-is if exists)
+            - Hugging Face model/dataset ID (e.g., "bert-base-uncased")
+            - ModelScope model/dataset ID
+            - URL-prefixed ID (e.g., "hf://model-name", "modelscope://model-name").
+              The prefix will override the platform argument.
+        repo_type (str, optional): Type of repository to download. Defaults to "model".
+            Common values include "model" and "dataset".
+        platform (Literal["hf", "huggingface", "modelscope"], optional):
+            Platform to download from. Defaults to "hf". Options:
+            - "hf" or "huggingface": Hugging Face Hub
+            - "modelscope": ModelScope platform
+        **kwargs: Additional arguments passed to the underlying download functions.
+
+    Returns:
+        str: Local path to the repository (either existing local path or downloaded cache path).
+
+    Raises:
+        FileNotFoundError: If the repository cannot be found or downloaded from any platform.
+        ValueError: If an unsupported platform is specified.
+        ImportError: If required dependencies for the specified platform are not installed.
+
+    Examples:
+        >>> # Local path (returned as-is)
+        >>> resolve_repo_path("/path/to/local/model")
+        "/path/to/local/model"
+
+        >>> # Hugging Face model
+        >>> resolve_repo_path("bert-base-uncased")
+        "/home/user/.cache/huggingface/hub/models--bert-base-uncased/..."
+
+        >>> # ModelScope model with explicit platform
+        >>> resolve_repo_path("damo/nlp_bert_backbone_base_std", platform="modelscope")
+        "/home/user/.cache/modelscope/hub/damo/nlp_bert_backbone_base_std/..."
+
+        >>> # URL-prefixed repository ID
+        >>> resolve_repo_path("hf://microsoft/DialoGPT-medium")
+        "/home/user/.cache/huggingface/hub/models--microsoft--DialoGPT-medium/..."
+    """
+    # If it's a HuggingFace Hub model id, download snapshot
+    if repo_id.startswith("hf://") or repo_id.startswith("huggingface://"):
+        repo_id = repo_id.replace("hf://", "").replace("huggingface://", "")
+        platform = "hf"
+    # If it's a ModelScope model id, download snapshot
+    elif repo_id.startswith("modelscope://"):
+        repo_id = repo_id.replace("modelscope://", "")
+        platform = "modelscope"
+
+    # If it's a local file or directory, return as is
+    if os.path.exists(repo_id):
+        return repo_id
+
+    try:
+        validate_and_suggest_corrections(platform, AVAILABLE_PLATFORMS)
+        # This will download the model to the cache and return the local path
+        if platform in ["hf", "huggingface"]:
+            local_path = huggingface_snapshot_download(
+                repo_id=repo_id, repo_type=repo_type, **kwargs
+            )
+        elif platform == "modelscope":
+            local_path = modelscope_snapshot_download(
+                repo_id=repo_id, repo_type=repo_type, **kwargs
+            )
+        else:
+            raise ValueError(
+                f"Unsupported platform: {platform}. Supported platforms are 'hf', 'huggingface', and 'modelscope'."
+            )
+        return local_path
+    except Exception as e:
+        raise FileNotFoundError(f"Could not resolve checkpoint: {repo_id}. Error: {e}")

fusion_bench/utils/state_dict_arithmetic.py

@@ -4,6 +4,7 @@ from typing import Callable, Dict, List, Literal, Union, cast
 
 import torch
 from torch import Tensor
+from tqdm.auto import tqdm
 
 from .parameters import check_parameters_all_equal
 from .type import BoolStateDictType, StateDictType
@@ -124,7 +125,11 @@ def state_dict_sub(
 
 
 def state_dict_add(
-    a: StateDictType, b: StateDictType, strict: bool = True, device=None
+    a: StateDictType,
+    b: StateDictType,
+    strict: bool = True,
+    device=None,
+    show_pbar: bool = False,
 ):
     """
     Returns the sum of two state dicts.
@@ -140,10 +145,10 @@ def state_dict_add(
     ans = {}
     if strict:
         check_parameters_all_equal([a, b])
-        for key in a:
+        for key in tqdm(tuple(a.keys())) if show_pbar else a:
             ans[key] = a[key] + b[key]
     else:
-        for key in a:
+        for key in tqdm(tuple(a.keys())) if show_pbar else a:
             if key in b:
                 ans[key] = a[key] + b[key]
     if device is not None:
@@ -175,7 +180,7 @@ def state_dict_mul(state_dict: StateDictType, scalar: float):
     return diff
 
 
-def state_dict_div(state_dict: StateDictType, scalar: float):
+def state_dict_div(state_dict: StateDictType, scalar: float, show_pbar: bool = False):
     """
     Returns the division of a state dict by a scalar.
 
@@ -187,7 +192,7 @@ def state_dict_div(state_dict: StateDictType, scalar: float):
         Dict: The division of the state dict by the scalar.
     """
     diff = OrderedDict()
-    for k in state_dict:
+    for k in tqdm(tuple(state_dict.keys())) if show_pbar else state_dict:
         diff[k] = state_dict[k] / scalar
     return diff
 

{fusion_bench-0.2.19.dist-info → fusion_bench-0.2.20.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: fusion_bench
-Version: 0.2.19
+Version: 0.2.20
 Summary: A Comprehensive Benchmark of Deep Model Fusion
 Author-email: Anke Tang <tang.anke@foxmail.com>
 License: MIT License
@@ -72,6 +72,14 @@ FusionBench is a benchmark suite designed to evaluate the performance of various
 
 Projects based on FusionBench and news from the community (descending order of date. If you have any work based on FusionBench, please feel free to let us know, we are willing to add it to the list. :partying_face:):
 
+<details>
+  <summary>The-Hai Nguyen, Dang Huu-Tien, Takeshi Suzuki, and Le-Minh Nguyen. RegMean++: Enhancing Effectiveness and Generalization of Regression Mean for Model Merging. Aug, 2025. https://www.arxiv.org/abs/2508.03121</summary>
+
+Regression Mean (RegMean), an approach that formulates model merging as a linear regression problem, aims to find the optimal weights for each linear layer in the merge model by minimizing the discrepancy in predictions between the merge and candidate models. RegMean provides a precise closed-form solution for the merging problem; therefore, it offers explainability and computational efficiency. However, RegMean merges each linear layer independently, overlooking how the features and information in the earlier layers propagate through the layers and influence the final prediction in the merge model. In this paper, we introduce RegMean++, a simple yet effective alternative to RegMean, that explicitly incorporates both intra- and cross-layer dependencies between merge models' layers into RegMean's objective. By accounting for these dependencies, RegMean++ better captures the behaviors of the merge model. Extensive experiments demonstrate that RegMean++ consistently outperforms RegMean across diverse settings, including in-domain (ID) and out-of-domain (OOD) generalization, sequential merging, large-scale tasks, and robustness under several types of distribution shifts. Furthermore, RegMean++ achieves competitive or state-of-the-art performance compared to various recent advanced model merging methods. 
+
+  <img width="1000" alt="image" src="docs/algorithms/images/regmean_vs_regmean_plusplus.png">
+</details>
+
 <details>
   <summary>Hao Mark Chen, et al. FW-Merging: Scaling Model Merging with Frank-Wolfe Optimization. Mar 2025. https://arxiv.org/abs/2503.12649</summary>
 

{fusion_bench-0.2.19.dist-info → fusion_bench-0.2.20.dist-info}/RECORD

@@ -43,12 +43,12 @@ fusion_bench/dataset/llama/stanford_shp.py,sha256=6ueXKnFXIBBobacU1h5WxGLZrSOtBk
 fusion_bench/dataset/llama/ultrachat.py,sha256=Go7WvrDAYnm184fdazHGRYLbSY6Xd7jrESyQeUJtOww,1736
 fusion_bench/dataset/llama/wikitext.py,sha256=9ZHR-nMfXRumd3o-PIj3n7B83YlVeqpGkZ2zJs2B-9Y,2883
 fusion_bench/dataset/llama/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-fusion_bench/method/__init__.py,sha256=TMELBu1IdKN86Id1rlNlr-vqsdArti_6mlKLfobHoL4,7888
+fusion_bench/method/__init__.py,sha256=o8t4R4fSTVaYuv2c0uD0_5ZolGuRk3nLawanYZghoCk,8015
 fusion_bench/method/base_algorithm.py,sha256=UuITuGnSskcKEwUVINuPoWJUwqGm9AIgyQIOCu8BMks,1162
 fusion_bench/method/dummy.py,sha256=hb1y6LR_geRZ5eRgGwt5zJUcHYorCeIbs5i76CvurUc,1031
 fusion_bench/method/ensemble.py,sha256=rGxvJTeorfcBuE_e0XO-0-MAc9un7ZCC46ikKGuAcN4,3077
 fusion_bench/method/model_recombination.py,sha256=2tviqmYSPOL0_Ktv8_gt_YzQ4tyCANHxXquUot_3Cgo,5360
-fusion_bench/method/simple_average.py,sha256=A_VYtHhECcxY0_Mppe5ThOY-ip6XUvvtPHsaQKSmDPc,4971
+fusion_bench/method/simple_average.py,sha256=omUIZn7VnrerAHbo9lNGUlL5EcpUeGv3C-fJczcXaOk,5458
 fusion_bench/method/ada_svd/__init__.py,sha256=4XzQbbvE9HI3NtEmEFvo8iC3ds_85vJXe7P7qJfL7kk,77
 fusion_bench/method/ada_svd/clip_vision.py,sha256=XvXgIdlShAREMsubRgphyycGrhWqSnuVBo6S9bNYSd0,12581
 fusion_bench/method/adamerging/__init__.py,sha256=nt0saBT_3bqghk-pINQ-XCWm9UWwSZllu4R1sDuAJAA,376
@@ -121,7 +121,7 @@ fusion_bench/method/linear/__init__.py,sha256=ChfkoOEAb-rUKwpowFPel-a1hRfS8gCrbn
 fusion_bench/method/linear/expo.py,sha256=LCHTWlsPm1Mjhrq0mfpWLVC7skkI9ZksGduy3TxULoU,3939
 fusion_bench/method/linear/linear_interpolation.py,sha256=IONw9BPiRJouY8bE9Abfyz7qVI_1B1n8KGZa0f7Pza8,2157
 fusion_bench/method/linear/llama_expo.py,sha256=ccECjhAqcFmzOIDyZ7e_aPzTM2Kj8u2D8TJytyz18YM,8476
-fusion_bench/method/linear/simple_average_for_llama.py,sha256=OcjvfG5nuUzdo3P4Xi1mO6ApRu51YAUYXG5lAMeD6rg,2711
+fusion_bench/method/linear/simple_average_for_llama.py,sha256=ZbkTEgTNYBq8p3q_cxA6gky0mdZs7FaGfHlZKfFEbiQ,2952
 fusion_bench/method/linear/task_arithmetic_for_llama.py,sha256=4SZpiTD7OzhWUXtcdK3PYdXbBGyDqiZd7oZOQ0lraN0,1963
 fusion_bench/method/lm_finetune/__init__.py,sha256=IFGAqXujX3Fabzl_tC6zZyOyPFJfVziL0qFtj5MVxj0,149
 fusion_bench/method/lm_finetune/bradley_terry_rm.py,sha256=ys_td1IeL3bzPTE0Cixlj2JooCaB7qseRwSDwroAk5A,18777
@@ -183,6 +183,9 @@ fusion_bench/method/regmean/__init__.py,sha256=VVqAkdHkb005Sc2XmeiedQYzb3q5aQNI8
 fusion_bench/method/regmean/clip_regmean.py,sha256=xhT7dYSCg9sPLL5ZUCCtcA-Ypw4PBHsOivrnz-3fDso,4931
 fusion_bench/method/regmean/gpt2_regmean.py,sha256=p2D3E8YAZsltsI6GM474UWNqPZfBqihLZ93ZLUpOJ_c,5565
 fusion_bench/method/regmean/regmean.py,sha256=IqkweSS4WXGbMquWaHWHijiZ6rMIaHfFSje8C8bbb6g,16587
+fusion_bench/method/regmean_plusplus/__init__.py,sha256=TD_KeTGXkbsh7tIsFDAnTDmYSvkc1ii2HfYQeNCu58A,141
+fusion_bench/method/regmean_plusplus/clip_regmean_plusplus.py,sha256=s0HMUHTg6MV8B_WnF6D8LHRhrbr9aHChxMhnDj8qweE,7385
+fusion_bench/method/regmean_plusplus/regmean_plusplus.py,sha256=and0YbngsWdq3zLBPOqTkhsCxR4WI9FjuoDQbLQfR4M,16069
 fusion_bench/method/slerp/__init__.py,sha256=Wgl9gg01Xou6jyZeBRD98kRnB_dAADDaPqRTHoONx9o,59
 fusion_bench/method/slerp/slerp.py,sha256=2_n10REnRoV5DuwCC0bDX8RM3MLL4Q_5rZiU0hICw2w,3406
 fusion_bench/method/slerp/slerp_utils.py,sha256=vksRo6n7FqY7By9aqbwTL4XV3BjcU_GrUl_r85Kpfjc,3504
@@ -255,7 +258,7 @@ fusion_bench/modelpool/nyuv2_modelpool.py,sha256=btuXmYxwfjI6MnGakhoOf53Iyb9fxYH
 fusion_bench/modelpool/causal_lm/__init__.py,sha256=F432-aDIgAbUITj4GNZS9dgUKKhaDMCbTeHB-9MecaQ,99
 fusion_bench/modelpool/causal_lm/causal_lm.py,sha256=7-mUWVGVsXyljH_06CmIyReClKx_xVjy5zeXTJcLQIk,8085
 fusion_bench/modelpool/clip_vision/__init__.py,sha256=3b9gN2bWUsoA1EmpitnIMnIlX7nklxbkn4WJ0QJtS2c,43
-fusion_bench/modelpool/clip_vision/modelpool.py,sha256=ADgzslXwYd95x42V26XvgS09WEKGfhH_AYuQmWKdT0w,5887
+fusion_bench/modelpool/clip_vision/modelpool.py,sha256=eemYxDPCovbdKo1yr96UJqieWxr7l_n0rQHsJT4JMno,7117
 fusion_bench/modelpool/openclip_vision/__init__.py,sha256=QDmAitKqUwRygN9QncdS_kGWZdfTKL4uUifC8xh9c10,47
 fusion_bench/modelpool/openclip_vision/modelpool.py,sha256=2MieB4PMvg85DaiYu49m3BzuBjib1xozJHTpYyHhRTs,11102
 fusion_bench/modelpool/seq2seq_lm/__init__.py,sha256=FnfSMHcwNHDQEMdB2HdK4WphQ6MufsRLUkczuALjM4Q,57
@@ -343,7 +346,7 @@ fusion_bench/programs/__init__.py,sha256=oGoRp2TMI6ELxyfkeTg2h27hZJEDz9x31Asmvwv
 fusion_bench/programs/base_program.py,sha256=0dX_KcMWASo53pr-ldzfUBWIjEXy6oeDWZBrfc7FIk8,195
 fusion_bench/programs/fabric_fusion_program.py,sha256=978t9Fw9kvw-Il7rJLR2jNI1OfSxkhq1c5-5D4BgnYU,13813
 fusion_bench/scripts/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-fusion_bench/scripts/cli.py,sha256=hw32XtmixFxYXwgAY7iRBMzma_XQjdf_FxPiXKL6dIc,1154
+fusion_bench/scripts/cli.py,sha256=gjNghi2VSW2gg41bn-WjzSkmTE3pDLoFk9c61zLJNso,1154
 fusion_bench/scripts/imgui.py,sha256=r9Glbfbwu3JCsX9TKQFwcHarvwA_G7ff0jWBUPW1S1U,7613
 fusion_bench/scripts/nyuv2_mtl_train.py,sha256=W1C45R9NdF4O-UjCx1bUxRTdFE0-FlRpwJHZ5gY18rI,3602
 fusion_bench/scripts/webui.py,sha256=ryA-2leSnHcYA88tTAYzJGDhiljbi0vl1Fibejzndlw,14398
@@ -419,14 +422,15 @@ fusion_bench/utils/instantiate_utils.py,sha256=57D8YP25OO-ArltOSsHDKtnNcA44m1yAq
 fusion_bench/utils/json.py,sha256=sVCqbm9mmyHybiui-O57KFt_ULrjLtN2wipSo6VDvqE,2533
 fusion_bench/utils/lazy_imports.py,sha256=v5l9cpHXPMaz1IVBmB5oOqefYr9vA3XvP340xT7Wy18,2796
 fusion_bench/utils/lazy_state_dict.py,sha256=Hu8PkhbJcUikXJxWUJ7vabu2uDbnUUF6UsRS0k8i71U,16841
-fusion_bench/utils/misc.py,sha256=Qc3_H8UMooOp81Ow89zqvM1sNPIybq1cbq7s4-4lsfU,1082
+fusion_bench/utils/misc.py,sha256=93q0m-HYWkPK91Co5lll_J0Dxs6YahW2lD_X8fUAyTk,2420
+fusion_bench/utils/modelscope.py,sha256=EZfvP6ExpagQXX2s0mbyE_yfmMHntQwvBN9taznnRmE,5705
 fusion_bench/utils/packages.py,sha256=L64paDi1SmeT3gRvRV6LaqB8AeGdzIYWIRI31qSQbSk,2110
 fusion_bench/utils/parameters.py,sha256=2vs8vo2o-nRA9NOMOYFye-X8-aHQZoYe54tM6n0r0RE,11757
 fusion_bench/utils/path.py,sha256=hRA1CPHNnTYBUmzbftH77sHvn4aTuybEK5Tth1skP-k,531
 fusion_bench/utils/pylogger.py,sha256=amlRsdqHpOjxmBl6f9TA8y0LaWelEWgQNcGgEGsVOIc,3333
 fusion_bench/utils/rich_utils.py,sha256=B8DhAYuVp23pG6ZnnYrUhcL-ikHZoQeTNqlM7u4pwwU,5786
 fusion_bench/utils/set.py,sha256=_43ZvGKJ_BK9sUslsSNhi7xEfuAQuyj3vViImnGpnCY,134
-fusion_bench/utils/state_dict_arithmetic.py,sha256=iz5YYhMJpg2-lBLBY8E1onV4i_GkRhJOGn2DjhLBbYE,11390
+fusion_bench/utils/state_dict_arithmetic.py,sha256=7iWb9cy-eYJgiltGJXnVCrLoNHfIMXZEL_AOg6OY0dw,11613
 fusion_bench/utils/tensorboard.py,sha256=9fkgNYR9LM38nPNkudcxL9TjLUseW-280M0k2nLff7o,1669
 fusion_bench/utils/timer.py,sha256=RC2hP8JqaibdL0FnRyUCBRf4m7CXyfn5tE16zBWZ7hg,1338
 fusion_bench/utils/type.py,sha256=2iu8PQzSzI2KopYwg4Pay7qpq7s_LKkl6Rhj-tjG3u0,630
@@ -437,7 +441,7 @@ fusion_bench/utils/plot/token_notebook.py,sha256=bsntXf46Zz_RavTxNiB9c3-KvHw7LFw
 fusion_bench/utils/strenum/__init__.py,sha256=id9ORi1uXrDxhbmVxitJ1KDwLS4H3AAwFpaK5h1cQzw,8531
 fusion_bench/utils/strenum/_name_mangler.py,sha256=o11M5-bURW2RBvRTYXFQIPNeqLzburdoWLIqk8X3ydw,3397
 fusion_bench/utils/strenum/_version.py,sha256=6JQRo9LcvODbCOeVFYQb9HNJ_J9XiG_Zbn8ws2A3BV8,18466
-fusion_bench-0.2.19.dist-info/licenses/LICENSE,sha256=nhnOJlw4CPuPVE0qvkGmxfFgHmKi-6nzXvTu8t0NUdg,1066
+fusion_bench-0.2.20.dist-info/licenses/LICENSE,sha256=nhnOJlw4CPuPVE0qvkGmxfFgHmKi-6nzXvTu8t0NUdg,1066
 fusion_bench_config/README.md,sha256=Lc8YSBJ5oxf9KV5kKDivJ9LRyGuraGQPmBbgbdVA-j4,703
 fusion_bench_config/clip-vit-base-patch32_robustness_corrupted.yaml,sha256=7IxLQoLRz-sRWyV8Vqc5kQcmYE_9YQz2_77pmvAkum8,1207
 fusion_bench_config/fabric_model_fusion.yaml,sha256=YwJx_aUXm4ca4_mVItKVUOesMvmBBRGudQIOqgc1EP8,974
@@ -622,9 +626,10 @@ fusion_bench_config/method/randes/superposed_model_soup.yaml,sha256=7M9qV_wCgrE3
 fusion_bench_config/method/randes/superposed_task_arithmetic.yaml,sha256=Pw0pZtwoMIPiqHfFNbN8wqNDyYb4L5p6fIOaaDSzJQg,498
 fusion_bench_config/method/randes/superposed_task_arithmetic_lora.yaml,sha256=xH8IkGnjvKLEWsms64toWhOrKIJG9dYfqQGOsVT4GDc,539
 fusion_bench_config/method/rankone_moe/rankone_moe.yaml,sha256=rYas_GFFHvn3AgKNrI0Zp4ElL9e3SppGPrFAMa_u9r8,863
-fusion_bench_config/method/regmean/clip_regmean.yaml,sha256=dxSJMRam6YMks7zYx4ACgvrLP5cndxzraVO93SGhyYo,425
+fusion_bench_config/method/regmean/clip_regmean.yaml,sha256=QfkCHCLK9wbyB1Tq1S7YT3351MbWzOjUQiALE-EJBgw,426
 fusion_bench_config/method/regmean/gpt2_regmean.yaml,sha256=n94aTboDdwSA7Tki8l_o8tYQkhXxPV8lRf-dRNPIsOs,422
 fusion_bench_config/method/regmean/regmean.yaml,sha256=ZgVVLx-lHwVgjtjTl4VZUlthh8yyua87QvoJfmNHud4,101
+fusion_bench_config/method/regmean_plusplus/clip_regmean_plusplus.yaml,sha256=A034ryEwvosqyQzA3KWs7kdp-3CUnoJtCujVywV-uzA,434
 fusion_bench_config/method/slerp/slerp.yaml,sha256=xldDUULtfCdwzAkQUb0C8-TmbW7FqcAlIOsPX8p4n6w,116
 fusion_bench_config/method/smile_upscaling/singular_projection_merging.yaml,sha256=ZMn_ImRjjc2uozf7ocQIzbgvFDpBV7S-34KptbBXVGo,200
 fusion_bench_config/method/smile_upscaling/smile_mistral_upscaling.yaml,sha256=VFMrkbO69d0wCjTQCuKysYGVe6hEwNu792g1QkhU5Mk,383
@@ -760,21 +765,21 @@ fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TA8_lora
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TA8_model_only.yaml,sha256=4_fQ7O6vUzIxLe-3mfY6qapx1rg5icQe_ODCbMspVRU,236
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL14.yaml,sha256=jBFj9O84KGWOwAI8a_In3Cq_C2caNE7JPYxhKaVDjsE,508
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL14_model_only.yaml,sha256=fvebppOvcZkVQfsCRllZzEU3ifbes4y7PHqS_gy0SYY,384
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL20.yaml,sha256=flHk78v6C4YzuT7pBJyQ0TT4U_ALXT9wWZb1iQnMPBk,508
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL20_model_only.yaml,sha256=qnps2pbmcRbWEMZv506cspQ2Fuerj_1dsZtnsccD0RQ,384
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL20.yaml,sha256=oPsGZjPxfb9wqmKzZv9MFeZNCtTh1AOY-01GK5fxMA4,2454
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_TALL20_model_only.yaml,sha256=pYoIPL2ebZGv7_dUQZQCDFVlU68K_VqUHQU0asYccmA,1384
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_individual.yaml,sha256=-jdB4ctj_NJcgbdYgog2-cighUSufAfpUvMOrGDblog,547
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch16_individual_lora.yaml,sha256=si8oNdCzCE_UDsHevyPGXxcJ0IZfRWHrrykHryazIvY,435
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8.yaml,sha256=1oY5QXG7flQq1990eN8GyjS1kak-dLPZIusV_dYyJVo,250
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8_control_task.yaml,sha256=5-g4DRsD5RBr8As6mDD9V8pT1JxKKPIUjwK8O7vBOMQ,688
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8_model_only.yaml,sha256=NdefoS2vRA4byRgjwKh8OplRJB48vg3jaeYiSjwQfAQ,108
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL10.yaml,sha256=uAVyNQQT16-oVB9CvxBNyoArni-giZDqkDPQogeJnP0,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL12.yaml,sha256=K5AWiFweGQ9CxQ0Wk-1ucfAs25WFFcWtpFEiio521rI,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL14.yaml,sha256=cmHBQY4zl4ySRJ5n11MIOzLERjPyxaFk1DGWFlrn-zQ,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL14_model_only.yaml,sha256=_2t8TV9T7BAiQda5fKCRIryTPg3XHYEM3hQUJqD1lXs,256
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL16.yaml,sha256=7Z4UVw3ManuEOw1D242UHn9uOv0LqUh2-uZt1M3718E,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL18.yaml,sha256=gGVJbI9LqenrGlFL3OCIqUxpf8IGM5GaXajgi9qVe1Y,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL20.yaml,sha256=V-p1JLhNwILgR0F4i6l8oOEQvZcYxs3J0Ly0VeAJY48,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL20_model_only.yaml,sha256=xYr0g5mdv0wly5HkTcnLq5yG6Mjj78XB7fGaCTk5KEc,256
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8.yaml,sha256=FUJYnr8pRVYEKoPLGrgGf6FH7ctbpCntG1YSYQV130Q,1013
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8_control_task.yaml,sha256=5nq8kWxu6rc5LJ83yVtzbTqaaLol_lymkbePVmGFo94,667
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TA8_model_only.yaml,sha256=vz1rmOT-bt56eDmWLT21NFq-Si_qE8EHcbWWi67VbGE,568
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL10.yaml,sha256=rwbWzJnWmVNiwxfyNFDVc-5neIZToQm9N1VZLBt55Kk,1290
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL12.yaml,sha256=vSiHAIRJI-p7V8u2BS45f20AfIPccjFZ2oImGUFvytg,1742
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL14.yaml,sha256=VfrDzBIDRZcBXl1PwRiB04HcvwEARAIweHvlGGNhNmk,1902
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL14_model_only.yaml,sha256=xh2lwb8sKMqdd6VZNj1lA2r063XXtQG0vUCXv76slNQ,1048
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL16.yaml,sha256=qHUHlJCfYm1oxgb4gzaS0BBQ5akmdAO8g1kvO2FPxCw,2066
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL18.yaml,sha256=383zCvjWgs5VmwBKWXcHfqqO8Fk75tD7IyhcvJ590ZU,2270
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL20.yaml,sha256=YRv9FFqWnY4yOEtEAL91TRf2xgPB8P7Y0CAOgtX31Kk,2454
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_TALL20_model_only.yaml,sha256=gEHHb4LVHUteJ-MKlupbZXktff6_wvGVL0CLv4cfrk0,1385
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_cars_and_dtd.yaml,sha256=V93v7cjxF0ZPJj0wX76Q-hSNvolUaTtoeWuAImSU53g,524
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_generalization_exp1.yaml,sha256=2WtCV1cJEEK3R-t4Tf-YB1AIZl-d0FkE6C0CsUBm9fw,625
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_generalization_exp2.yaml,sha256=BmQ0JP8Oa5_V5pJ55nJS1xR-OIPmvySSqQ36l2jAB1w,625
@@ -782,20 +787,20 @@ fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_individu
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_mtl.yaml,sha256=pQr8lF-hIFrPXPcZYlbSxx8PF8EClZ6nm2L4yqEmHTk,176
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_robustness_clean.yaml,sha256=7oQtoqXs37fctajb6E7UOB0GT515eEGzFNm93dWOKKk,509
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_robustness_corrupted.yaml,sha256=txMh1k0O3Spusqewp7zV0N0L9e2fg87lviDEnNJSHGQ,900
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_single_finetuned.yaml,sha256=SBTyUX3wJwzdCTvZsW14FqaQP5r_nHPvusggGzP9P4o,148
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_single_task_projection.yaml,sha256=urlcrY5TEDOFJqYYmbaIY2Mi6_jIRdECnqo1gXWnPyU,390
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_and_cars.yaml,sha256=kC_DbHgZoC6p2-26e-jtjMS9mxyHMT-_B684UNQ59vo,533
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_and_dtd.yaml,sha256=BeF2ygrcElkvPlUo9LV9XxBO1Y75XxDSSS52cU-gNq4,503
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_cars_and_dtd.yaml,sha256=Dl08CAHcqbUPZkOYTAycJ_clkAPvkDSpxPxsY0uz54o,591
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_svhn_and_mnist.yaml,sha256=fYthV8iwRvF-b4-OCIFW1Rud-BVoLx4Oo3DzVszfqek,175
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_two_tasks_control_task.yaml,sha256=5uw3lD-bdHNQ76osDb0SBnzsdWABw08HYtUkDG-jioI,477
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TA8.yaml,sha256=-Tt_YggxkuIGT4_q5FR16zPvW2wWhGJ5LL8omxvHjvw,380
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TA8_model_only.yaml,sha256=PrDQr04UnhAciDrdtUutx-prMxF1Cs4jrEar7uJ-1Es,238
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_single_finetuned.yaml,sha256=nm22fkXYUGAgZ39N2AzSE4CUVaODHgi8HTIxBWuTz3M,240
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_single_task_projection.yaml,sha256=i78xIL-vP28dYZaXntLsm7e9IdI2yAeUwZZze5fd9Do,288
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_and_cars.yaml,sha256=gDEzNfwsMtIu2xH8WSIUblx4ZyL1FsLXoSEpXPHiMaI,482
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_and_dtd.yaml,sha256=Ej1NOsPJzLVZmBI9jK3QMhZg198IqThUZwt8_6GizUM,442
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_sun397_cars_and_dtd.yaml,sha256=V1KSVfjY0hqbJFnnOkoe7SD3qKBWmYyx2fqrLTX05bE,548
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_svhn_and_mnist.yaml,sha256=jOTbHpyqBVepW7KhorStMjH9cVzHZvq6K7osgYnsVIU,443
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-base-patch32_two_tasks_control_task.yaml,sha256=dFVylgWlqWWW_Hh0N7b1m5A8WYK-r-tO7TxEviR5dCY,382
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TA8.yaml,sha256=ZuH1oweRcl08OeXNIxsb68PxYwdrUfliaHYa4s64ljo,939
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TA8_model_only.yaml,sha256=LmILwWINU8cRVvxx3IKZPFcAzPyMnr8OlCJ0TDs1WMg,573
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL14.yaml,sha256=TyF7CKXpBOiDoLtDVvZuBzPI9gEJo_c99j3Gwkl3FWw,510
 fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL14_model_only.yaml,sha256=i5XGxa2FoW2464R4k5AG-7r5qmzjHXkCSm1Om2cshik,386
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL20.yaml,sha256=yC2U_IoBAhawgSahY_mdi7ea5kJ2SSRPJ2FM-bA-E9M,510
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL20_model_only.yaml,sha256=a2nviqKSRNoQScYVbj5buq0PbUzmYJwNWdPBUoLaeV8,386
-fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_individual.yaml,sha256=G6yvZuWOKb75RLn6tu2LPnwHUyvoxPfL_wqb_B11aZo,549
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL20.yaml,sha256=FuPWQbC9xEV5wZjuo835gOMNgbzmpK9RbjFjA_HOzqo,2476
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_TALL20_model_only.yaml,sha256=9PCkbrNnQSKTsm4eoUvVgjGd3IY7wHBC4LWj4kOdY4Y,1406
+fusion_bench_config/modelpool/CLIPVisionModelPool/clip-vit-large-patch14_individual.yaml,sha256=bqnyzgwIvDtV3Fb-uLf9mdFv0NW1C392lxGsGUPLsKE,400
 fusion_bench_config/modelpool/CausalLMPool/Qwen2.5-1.5B_math_and_coder.yaml,sha256=HZXjqbZKpSZCHb-G8qjj03PcvXg_8mrAuewDHZp0oEw,263
 fusion_bench_config/modelpool/CausalLMPool/deepseek-v2-lite.yaml,sha256=8gr8ZtgegSHV0GHtJBiEgdYbRe8UHhO4_y8dayxZChk,506
 fusion_bench_config/modelpool/CausalLMPool/llama_alpaca_cleaned.yaml,sha256=oDsZkuAoh1mWUC7jZNzw8794zgX2bV5Z0esXpvbTs-c,643
@@ -877,8 +882,8 @@ fusion_bench_config/taskpool/LMEvalHarnessTaskPool/lm_eval.yaml,sha256=3q-KMuFaM
 fusion_bench_config/taskpool/OpenCLIPVisionModelTaskPool/ViT-B-16_TA8.yaml,sha256=GjpiiRownrBCpl-TNwWRW2PYePbF-Cl99jlLNPrK5T4,1017
 fusion_bench_config/taskpool/OpenCLIPVisionModelTaskPool/ViT-B-32_TA8.yaml,sha256=WwiYMQKehtJixDPnu5o3vcWe4yJksXTWRqOzm3uVWXQ,1017
 fusion_bench_config/taskpool/OpenCLIPVisionModelTaskPool/ViT-L-14_TA8.yaml,sha256=xGRt0J9joXTzWUew6DvoYprAWlPXhaVFw5AX4im5VQw,1017
-fusion_bench-0.2.19.dist-info/METADATA,sha256=5pl4dtlAYklMMiMLBeKNaHqCQRd7sLSct7aIh9JIoGY,21966
-fusion_bench-0.2.19.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-fusion_bench-0.2.19.dist-info/entry_points.txt,sha256=iUQ8MCJvda7HP4vYh2n1Teoapb4G9PBVYZkAfcc5SHU,116
-fusion_bench-0.2.19.dist-info/top_level.txt,sha256=BuO4TL6iHL_2yPBUX9-LlIrHRczA_BNMIFwweK0PQEI,13
-fusion_bench-0.2.19.dist-info/RECORD,,
+fusion_bench-0.2.20.dist-info/METADATA,sha256=fa8CMvPeD8fg_R8YuvjM_AggXMAVsqg63fZy3I4TMmc,23634
+fusion_bench-0.2.20.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+fusion_bench-0.2.20.dist-info/entry_points.txt,sha256=iUQ8MCJvda7HP4vYh2n1Teoapb4G9PBVYZkAfcc5SHU,116
+fusion_bench-0.2.20.dist-info/top_level.txt,sha256=BuO4TL6iHL_2yPBUX9-LlIrHRczA_BNMIFwweK0PQEI,13
+fusion_bench-0.2.20.dist-info/RECORD,,

fusion_bench_config/method/regmean/clip_regmean.yaml

@@ -5,7 +5,7 @@ exclude_param_names_regex: []
 num_regmean_examples: 256
 weight_transpose: true
 # float, reduce non-diagonal elements in regmean weights by multiplying this scalar
-reduce_non_diagonal_ratio: 0.6
+reduce_non_diagonal_ratio: 0.95
 dataloader_kwargs:
   batch_size: 32
   num_workers: 0

fusion_bench_config/method/regmean_plusplus/clip_regmean_plusplus.yaml

@@ -0,0 +1,11 @@
+_target_: fusion_bench.method.RegMeanAlgorithmForCLIPPlusPlus
+# list, regular expression of names of parameters that need to be excluded
+exclude_param_names_regex: []
+# numbers of examples to compute regmean weights
+num_regmean_examples: 256
+weight_transpose: true
+# float, reduce non-diagonal elements in regmean weights by multiplying this scalar
+reduce_non_diagonal_ratio: 0.95
+dataloader_kwargs:
+  batch_size: 32
+  num_workers: 0