whenxuan: init

.gitignore

@@ -0,0 +1,210 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+.vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# The model params
+.safetensors

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright wwhenxuan
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

README.md

@@ -0,0 +1,2 @@
+# SymTime-NeurIPS2025-Huggingface
+The pipeline and model config of SymTime model for Huggingface

config.json

@@ -0,0 +1,24 @@
+
+{
+  "architectures": [
+      "SymTimeModel"
+    ],
+  "_name_or_path": "FlowVortex/SymTime",
+  "auto_map": {
+    "AutoConfig": "configuration_symtime.SymTimeConfig",
+    "AutoModel": "model.SymTimeModel"
+  },
+  "patch_size": 16,
+  "num_layers": 6,
+  "d_model": 512,
+  "d_ff": 2048,
+  "num_heads": 8,
+  "norm": "BatchNorm",
+  "dropout": 0.1,
+  "act": "gelu",
+  "pre_norm": false,
+  "initializer_factor": 0.05,
+  "model_type": "time_series_transformer",
+  "torch_dtype": "float32",
+  "transformers_version": "5.5.4"
+}

configuration_symtime.py

@@ -0,0 +1,64 @@
+from dataclasses import dataclass
+from typing import List, Literal, Optional, Dict
+from enum import Enum
+
+from transformers.configuration_utils import PretrainedConfig
+
+
+@dataclass
+class SymTimeConfig(PretrainedConfig):
+    """
+    Time series encoder configuration for SymTime Model.
+
+    Parameters
+    -----------
+    patch_size
+        The size of the patch to be used for the input data.
+    num_layers
+        The number of layers to be used for the encoder.
+    d_model
+        The dimension of the model.
+    d_ff
+        The dimension of the feedforward network.
+    num_heads
+        The number of heads to be used for the attention mechanism.
+    norm
+        The normalization to be used for the encoder.
+    attn_dropout
+        The dropout rate to be used for the attention mechanism.
+    dropout
+        The dropout rate to be used for the encoder.
+    act
+        The activation function to be used for the encoder.
+    pre_norm
+        Whether to use pre-norm for the encoder.
+    """
+
+    model_type = "time_series_transformer"
+
+    def __init__(
+        self,
+        patch_size: int = 16,
+        num_layers: int = 6,
+        d_model: int = 512,
+        d_ff: int = 2048,
+        num_heads: int = 8,
+        norm: str = "BatchNorm",
+        dropout: float = 0.1,
+        act: str = "gelu",
+        pre_norm: bool = False,
+        initializer_factor: float = 0.05,
+        **kwargs,
+    ) -> None:
+        self.patch_size = patch_size
+        self.num_layers = num_layers
+        self.d_model = d_model
+        self.num_heads = num_heads
+        self.d_ff = d_ff
+        self.norm = norm
+        self.dropout = dropout
+        self.act = act
+        self.pre_norm = pre_norm
+        self.initializer_factor = initializer_factor
+
+        super().__init__(**kwargs)

layers.py

@@ -0,0 +1,401 @@
+from typing import Optional, Union, Tuple, Callable
+import math
+
+import numpy as np
+import torch
+from torch import nn
+from torch import Tensor
+import torch.nn.functional as F
+from einops import rearrange
+
+
+def get_activation_fn(activation: Union[str, Callable]) -> nn.Module:
+    """Select the activation function to use."""
+    if callable(activation):
+        return activation()
+    elif activation.lower() == "relu":
+        return nn.ReLU()
+    elif activation.lower() == "gelu":
+        return nn.GELU()
+    raise ValueError(
+        f'{activation} is not available. You can use "relu", "gelu", or a callable'
+    )
+
+
+class Transpose(nn.Module):
+    """Transpose the dimensions of the input tensor"""
+
+    def __init__(self, *dims, contiguous=False) -> None:
+        super().__init__()
+        self.dims, self.contiguous = dims, contiguous
+
+    def forward(self, x: Tensor) -> Tensor:
+        if self.contiguous:
+            return x.transpose(*self.dims).contiguous()
+        else:
+            return x.transpose(*self.dims)
+
+
+class PositionalEmbedding(nn.Module):
+    """Adding the positional encoding to the input for Transformer"""
+
+    def __init__(self, hidden_size: int, max_len: int = 5000) -> None:
+        super(PositionalEmbedding, self).__init__()
+
+        # Calculate the positional encoding once in the logarithmic space.
+        pe = torch.zeros(
+            max_len, hidden_size
+        ).float()  # Initialize a tensor of zeros with shape (max_len, hidden_size) to store positional encodings
+        pe.requires_grad = (
+            False  # Positional encodings do not require gradients as they are fixed
+        )
+
+        position = (
+            torch.arange(0, max_len).float().unsqueeze(1)
+        )  # Generate a sequence from 0 to max_len-1 and add a dimension at the 1st axis
+        div_term = (
+            torch.arange(0, hidden_size, 2).float() * -(math.log(10000.0) / hidden_size)
+        ).exp()  # Calculate the divisor term in the positional encoding formula
+
+        pe[:, 0::2] = torch.sin(
+            position * div_term
+        )  # Apply the sine function to the even columns of the positional encoding matrix
+        pe[:, 1::2] = torch.cos(
+            position * div_term
+        )  # Apply the cosine function to the odd columns of the positional encoding matrix
+
+        pe = pe.unsqueeze(
+            0
+        )  # Add a batch dimension, changing the shape to (1, max_len, hidden_size)
+        self.register_buffer(
+            "pe", pe
+        )  # Register the positional encodings as a buffer, which will not be updated as model parameters
+
+    def forward(self, x: Tensor) -> Tensor:
+        # Return the first max_len positional encodings that match the length of input x
+        return x + self.pe[:, : x.size(1)]
+
+
+class TSTEncoder(nn.Module):
+    """Time series encoder backbone of SymTime"""
+
+    def __init__(
+        self,
+        patch_size: int = 16,
+        num_layers: int = 3,
+        hidden_size: int = 128,
+        num_heads: int = 16,
+        d_k: int = None,
+        d_v: int = None,
+        d_ff: int = 256,
+        norm: str = "BatchNorm",
+        attn_dropout: float = 0.0,
+        dropout: float = 0.0,
+        act: str = "gelu",
+        store_attn: bool = False,
+        pre_norm: bool = False,
+    ) -> None:
+        super().__init__()
+        # The Linear layer to project the input patches to the model dimension
+        self.W_P = nn.Linear(patch_size, hidden_size)
+
+        # Positional encoding
+        self.pe = PositionalEmbedding(hidden_size=hidden_size)
+
+        # Residual dropout
+        self.dropout = nn.Dropout(dropout)
+
+        # Create the [CLS] token
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, hidden_size))
+        self.cls_mask = nn.Parameter(torch.ones(1, 1).bool(), requires_grad=False)
+
+        # Create the encoder layer of the model backbone
+        self.layers = nn.ModuleList(
+            [
+                TSTEncoderLayer(
+                    hidden_size=hidden_size,
+                    num_heads=num_heads,
+                    d_k=d_k,
+                    d_v=d_v,
+                    d_ff=d_ff,
+                    norm=norm,
+                    attn_dropout=attn_dropout,
+                    dropout=dropout,
+                    activation=act,
+                    pre_norm=pre_norm,
+                    store_attn=store_attn,
+                )
+                for _ in range(num_layers)
+            ]
+        )
+
+        # model params init
+        self.apply(self._init_weights)
+
+    def _init_weights(self, m: nn.Module) -> None:
+        """model params init through apply methods"""
+        if isinstance(m, nn.Linear):
+            nn.init.xavier_uniform_(m.weight)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    def forward(
+        self,
+        x: Tensor,  # x: [batch_size, patch_num, patch_size]
+        attn_mask: Optional[Tensor] = None,  # attn_mask: [batch, num_patch]
+        return_cls_token: bool = True,  # whether to return the CLS token
+    ) -> Tensor:
+        """ """
+        batch_size = x.size(0)
+
+        # Input patching embedding
+        x = self.W_P(x)  # x: [batch_size, patch_num, model_dim]
+
+        # Add the [CLS] token
+        cls_token = self.cls_token.expand(batch_size, -1, -1)
+        x = torch.cat([cls_token, x], dim=1)
+        # adjust the attn mask
+        if attn_mask is not None:
+            attn_mask = torch.cat(
+                [self.cls_mask.expand(batch_size, -1), attn_mask], dim=1
+            )
+
+        # Add the positional embedding
+        x = self.pe(x)
+        x = self.dropout(x)  # x: [batch_size, patch_num, hidden_size]
+
+        for mod in self.layers:
+            x = mod(x, attn_mask=attn_mask)
+
+        if not return_cls_token:
+            # If not returning the CLS token, remove it from the output
+            return x[:, 1:, :]
+
+        return x
+
+
+class TSTEncoderLayer(nn.Module):
+    """Patch-based Transformer module sublayer"""
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        d_k: int = None,
+        d_v: int = None,
+        d_ff: int = 256,
+        store_attn: int = False,
+        norm: str = "BatchNorm",
+        attn_dropout: float = 0.0,
+        dropout: float = 0.0,
+        bias: bool = True,
+        activation: str = "gelu",
+        pre_norm: bool = False,
+    ) -> None:
+        super(TSTEncoderLayer, self).__init__()
+
+        assert (
+            not hidden_size % num_heads
+        ), f"hidden_size ({hidden_size}) must be divisible by num_heads ({num_heads})"
+        # If not specified, the number of heads is divided
+        d_k = hidden_size // num_heads if d_k is None else d_k
+        d_v = hidden_size // num_heads if d_v is None else d_v
+
+        # Create the multi-head attention
+        self.self_attn = MultiHeadAttention(
+            hidden_size,
+            num_heads,
+            d_k,
+            d_v,
+            attn_dropout=attn_dropout,
+            proj_dropout=dropout,
+        )
+
+        # Add & Norm
+        self.dropout_attn = nn.Dropout(dropout)
+        if "batch" in norm.lower():
+            self.norm_attn = nn.Sequential(
+                Transpose(1, 2), nn.BatchNorm1d(hidden_size), Transpose(1, 2)
+            )
+        else:
+            self.norm_attn = nn.LayerNorm(hidden_size)
+
+        # Position-wise Feed-Forward
+        self.ff = nn.Sequential(
+            nn.Linear(hidden_size, d_ff, bias=bias),
+            get_activation_fn(activation),
+            nn.Dropout(dropout),
+            nn.Linear(d_ff, hidden_size, bias=bias),
+        )
+
+        # Add & Norm
+        self.dropout_ffn = nn.Dropout(dropout)
+        if "batch" in norm.lower():
+            self.norm_ffn = nn.Sequential(
+                Transpose(1, 2), nn.BatchNorm1d(hidden_size), Transpose(1, 2)
+            )
+        else:
+            self.norm_ffn = nn.LayerNorm(hidden_size)
+
+        # use pre-norm or not
+        self.pre_norm = pre_norm
+        self.store_attn = store_attn
+        self.attn = None
+
+    def forward(
+        self, src: Tensor, attn_mask: Optional[Tensor] = None
+    ) -> Union[Tuple[Tensor, Tensor], Tensor]:
+        """Multi-Head attention sublayer"""
+
+        # Whether to use pre-norm for attention layer
+        if self.pre_norm:
+            src = self.norm_attn(src)
+
+        # Multi-Head attention
+        src2, attn = self.self_attn(src, src, src, attn_mask=attn_mask)
+        if self.store_attn:
+            self.attn = attn
+
+        # Add: residual connection with residual dropout
+        src = src + self.dropout_attn(src2)
+        if not self.pre_norm:
+            src = self.norm_attn(src)
+
+        # Whether to use pre-norm for ffn layer
+        if self.pre_norm:
+            src = self.norm_ffn(src)
+
+        # Position-wise Feed-Forward
+        src2 = self.ff(src)
+
+        # Add: residual connection with residual dropout
+        src = src + self.dropout_ffn(src2)
+        if not self.pre_norm:
+            src = self.norm_ffn(src)
+
+        return src
+
+
+class MultiHeadAttention(nn.Module):
+    """Multi-head attention mechanism layer"""
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        d_k: int = None,
+        d_v: int = None,
+        attn_dropout: float = 0.0,
+        proj_dropout: float = 0.0,
+        qkv_bias: bool = True,
+    ) -> None:
+        """Multi Head Attention Layer
+        Input shape:
+            Q:       [batch_size (bs) x max_q_len x hidden_size]
+            K, V:    [batch_size (bs) x q_len x hidden_size]
+            mask:    [q_len x q_len]
+        """
+        super().__init__()
+        d_k = hidden_size // num_heads if d_k is None else d_k
+        d_v = hidden_size // num_heads if d_v is None else d_v
+
+        self.num_heads, self.d_k, self.d_v = num_heads, d_k, d_v
+
+        self.W_Q = nn.Linear(hidden_size, d_k * num_heads, bias=qkv_bias)
+        self.W_K = nn.Linear(hidden_size, d_k * num_heads, bias=qkv_bias)
+        self.W_V = nn.Linear(hidden_size, d_v * num_heads, bias=qkv_bias)
+
+        # Scaled Dot-Product Attention (multiple heads)
+        self.sdp_attn = _ScaledDotProductAttention(
+            hidden_size, num_heads, attn_dropout=attn_dropout
+        )
+
+        # Project output
+        self.to_out = nn.Sequential(
+            nn.Linear(num_heads * d_v, hidden_size), nn.Dropout(proj_dropout)
+        )
+
+    def forward(
+        self,
+        q: Tensor,
+        k: Optional[Tensor] = None,
+        v: Optional[Tensor] = None,
+        attn_mask: Optional[Tensor] = None,
+    ):
+        bs = q.size(0)
+        if k is None:
+            k = q
+        if v is None:
+            v = q
+
+        # Linear (+ split in multiple heads)
+        q_s = self.W_Q(q).view(bs, -1, self.num_heads, self.d_k).transpose(1, 2)
+        k_s = self.W_K(k).view(bs, -1, self.num_heads, self.d_k).permute(0, 2, 3, 1)
+        v_s = self.W_V(v).view(bs, -1, self.num_heads, self.d_v).transpose(1, 2)
+
+        # Apply Scaled Dot-Product Attention (multiple heads)
+        output, attn_weights = self.sdp_attn(q_s, k_s, v_s, attn_mask=attn_mask)
+
+        # back to the original inputs dimensions
+        output = (
+            output.transpose(1, 2).contiguous().view(bs, -1, self.num_heads * self.d_v)
+        )
+        output = self.to_out(output)
+
+        return output, attn_weights
+
+
+class _ScaledDotProductAttention(nn.Module):
+    r"""Scaled Dot-Product Attention module (Attention is all you need by Vaswani et al., 2017) with optional residual attention from previous layer
+    (Realformer: Transformer likes residual attention by He et al, 2020) and locality self sttention (Vision Transformer for Small-Size Datasets
+    by Lee et al, 2021)"""
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        attn_dropout: float = 0.0,
+        res_attention: bool = False,
+    ):
+        super().__init__()
+        self.attn_dropout = nn.Dropout(attn_dropout)
+        self.res_attention = res_attention
+        head_dim = hidden_size // num_heads
+        self.scale = nn.Parameter(torch.tensor(head_dim**-0.5), requires_grad=False)
+
+    def forward(
+        self, q: Tensor, k: Tensor, v: Tensor, attn_mask: Optional[Tensor] = None
+    ) -> Union[Tuple[Tensor, Tensor, Tensor], Tuple[Tensor, Tensor]]:
+        """
+        :param q: [batch_size, num_heads, num_token, d_k]
+        :param k: [batch_size, num_heads, d_k, num_token]
+        :param v: [batch_size, num_heads, num_token, d_k]
+        :param attn_mask: [batch_size, num_heads, num_token]
+        """
+
+        # Scaled MatMul (q, k) - similarity scores for all pairs of positions in an input sequence
+        attn_scores = torch.matmul(q, k) * self.scale
+
+        # Attention mask (optional)
+        if (
+            attn_mask is not None
+        ):  # attn_mask with shape [q_len x seq_len] - only used when q_len == seq_len
+            attn_mask = rearrange(attn_mask, "b i -> b 1 i 1") * rearrange(
+                attn_mask, "b i -> b 1 1 i"
+            )
+            if attn_mask.dtype == torch.bool:
+                attn_scores.masked_fill_(attn_mask, -np.inf)
+            else:
+                attn_scores += attn_mask
+
+        # normalize the attention weights
+        attn_weights = F.softmax(attn_scores, dim=-1)
+        attn_weights = self.attn_dropout(attn_weights)
+
+        # compute the new values given the attention weights
+        output = torch.matmul(attn_weights, v)
+
+        return output, attn_weights

model.py

@@ -0,0 +1,140 @@
+from typing import Tuple
+
+import torch
+import torch.nn as nn
+from torch import Tensor
+from torch.nn import functional as F
+from einops import rearrange, repeat
+from transformers.modeling_utils import PreTrainedModel
+
+from configuration_symtime import SymTimeConfig
+from layers import MultiHeadAttention, TSTEncoder, TSTEncoderLayer
+
+
+class SymTimeModel(PreTrainedModel):
+    """
+    SymTime Model for Huggingface.
+
+    Parameters
+    ----------
+    config: SymTimeConfig
+        The configuration of the SymTime model.
+
+    Attributes
+    ----------
+    config: SymTimeConfig
+        The configuration of the SymTime model.
+    encoder: TSTEncoder
+        The encoder of the SymTime model.
+
+    Methods
+    -------
+    forward(x: Tensor) -> Tuple[Tensor, Tensor]:
+        Forward pass of the SymTime model.
+
+    _init_weights(module: nn.Module) -> None:
+        Initialize weights for the SymTime encoder stack.
+    """
+
+    def __init__(self, config: SymTimeConfig):
+        super().__init__(config)
+        self.config = config
+        self.encoder = TSTEncoder(
+            patch_size=config.patch_size,
+            num_layers=config.num_layers,
+            hidden_size=config.d_model,
+            num_heads=config.num_heads,
+            d_ff=config.d_ff,
+            norm=config.norm,
+            attn_dropout=config.dropout,
+            dropout=config.dropout,
+            act=config.act,
+            pre_norm=config.pre_norm,
+        )
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def _init_weights(self, module) -> None:
+        """Initialize weights for the SymTime encoder stack.
+
+        The model is built on top of Hugging Face `PreTrainedModel`, so this method
+        is called recursively via `post_init()`. We keep the initialization aligned
+        with the current backbone structure in `layers.py`:
+
+        - `TSTEncoder.W_P`: patch projection linear layer
+        - `TSTEncoder.cls_token`: learnable CLS token
+        - `TSTEncoderLayer.self_attn`: Q/K/V and output projections
+        - `TSTEncoderLayer.ff`: feed-forward linear layers
+        - `LayerNorm` / `BatchNorm1d`: normalization layers
+        """
+        super()._init_weights(module)
+
+        factor = self.config.initializer_factor
+        d_model = self.config.d_model
+        num_heads = self.config.num_heads
+        d_k = d_model // num_heads
+        d_v = d_k
+
+        if isinstance(module, nn.Linear):
+            nn.init.normal_(
+                module.weight, mean=0.0, std=factor * (module.in_features**-0.5)
+            )
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+
+        elif isinstance(module, nn.LayerNorm):
+            nn.init.ones_(module.weight)
+            nn.init.zeros_(module.bias)
+
+        elif isinstance(module, nn.BatchNorm1d):
+            if module.weight is not None:
+                nn.init.ones_(module.weight)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+
+        elif isinstance(module, TSTEncoder):
+            if hasattr(module, "cls_token") and module.cls_token is not None:
+                nn.init.normal_(module.cls_token, mean=0.0, std=factor)
+            if hasattr(module, "W_P") and isinstance(module.W_P, nn.Linear):
+                nn.init.normal_(
+                    module.W_P.weight,
+                    mean=0.0,
+                    std=factor * (module.W_P.in_features**-0.5),
+                )
+                if module.W_P.bias is not None:
+                    nn.init.zeros_(module.W_P.bias)
+
+        elif isinstance(module, MultiHeadAttention):
+            nn.init.normal_(module.W_Q.weight, mean=0.0, std=factor * (d_model**-0.5))
+            nn.init.normal_(module.W_K.weight, mean=0.0, std=factor * (d_model**-0.5))
+            nn.init.normal_(module.W_V.weight, mean=0.0, std=factor * (d_model**-0.5))
+            if module.W_Q.bias is not None:
+                nn.init.zeros_(module.W_Q.bias)
+            if module.W_K.bias is not None:
+                nn.init.zeros_(module.W_K.bias)
+            if module.W_V.bias is not None:
+                nn.init.zeros_(module.W_V.bias)
+
+            out_proj = module.to_out[0]
+            nn.init.normal_(
+                out_proj.weight, mean=0.0, std=factor * ((num_heads * d_v) ** -0.5)
+            )
+            if out_proj.bias is not None:
+                nn.init.zeros_(out_proj.bias)
+
+        elif isinstance(module, TSTEncoderLayer):
+            for submodule in module.ff:
+                if isinstance(submodule, nn.Linear):
+                    nn.init.normal_(
+                        submodule.weight,
+                        mean=0.0,
+                        std=factor * (submodule.in_features**-0.5),
+                    )
+                    if submodule.bias is not None:
+                        nn.init.zeros_(submodule.bias)
+
+    def forward(
+        self, x: Tensor, return_cls_token: bool = True
+    ) -> Tuple[Tensor, Tensor]:
+        return self.encoder(x, return_cls_token=return_cls_token)

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00fd4239b3110418392f60f6b5fc93604b75bb7aad45b05be94e0397b3d81334
+size 85998105

requirements.txt

@@ -0,0 +1,3 @@
+einops
+torch
+transformers