diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.Allegro.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.Allegro.rst index 1174823165ca65ce89e856719f89d47d5e3abcee..d20c489b3665418d890b6c4ae220e3401d931575 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.Allegro.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.Allegro.rst @@ -1,13 +1,13 @@ mindchemistry.cell.Allegro ============================ -.. py:class:: mindchemistry.cell.Allegro(l_max: int = 1, parity_setting="o3_full", num_layers: int = 1, env_embed_multi: int = 8, avg_num_neighbor: float = 1.0, two_body_kwargs=None, latent_kwargs=None, env_embed_kwargs=None, irreps_in=None, enable_mix_precision=False) +.. py:class:: mindchemistry.cell.Allegro(l_max=1, parity_setting="o3_full", num_layers=1, env_embed_multi=8, avg_num_neighbor=1.0, two_body_kwargs=None, latent_kwargs=None, env_embed_kwargs=None, irreps_in=None, enable_mix_precision=False) Allegro网络。 参数： - **l_max** (int) - 球谐函数特征的最大阶数。默认值：``1``。 - - **parity_setting** (string) - 对称性相关设置。默认值：``"o3_full"``。 + - **parity_setting** (string) - 对称性相关设置。默认值：``"o3_full"`` 。 - **num_layers** (int) - Allegro 网络的层数。默认值：``1``。 - **env_embed_multi** (int) - 网络中特征的通道数。默认值：``8``。 - **avg_num_neighbor** (float) - 平均邻近原子数量。默认值：``1.0``。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.AttentionInteractionNetwork.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.AttentionInteractionNetwork.rst index 7778f3a537968f0a3a629102da109eae1e335619..7f76d093fa78605b8d4b072eb960f3e19c1a12dc 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.AttentionInteractionNetwork.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.AttentionInteractionNetwork.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.AttentionInteractionNetwork ================================================== -.. py:class:: mindchemistry.cell.orb.AttentionInteractionNetwork(num_node_in: int, num_node_out: int, num_edge_in: int, num_edge_out: int, num_mlp_layers: int, mlp_hidden_dim: int, attention_gate: str = "sigmoid", distance_cutoff: bool = True, polynomial_order: int = 4, cutoff_rmax: float = 6.0) +.. py:class:: mindchemistry.cell.orb.AttentionInteractionNetwork(num_node_in, num_node_out, num_edge_in, num_edge_out, num_mlp_layers, mlp_hidden_dim, attention_gate="sigmoid", distance_cutoff=True, polynomial_order=4, cutoff_rmax=6.0) 注意力交互网络。实现基于注意力机制的消息传递神经网络层，用于分子图的边更新。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.EnergyHead.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.EnergyHead.rst index fb549db328a7b26008669b6bbe43d7dc1a925bd9..fedcbfcb35d8e5d3b44f8ae1e0c0a0705fde7c79 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.EnergyHead.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.EnergyHead.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.EnergyHead ================================== -.. py:class:: mindchemistry.cell.orb.EnergyHead(latent_dim: int, num_mlp_layers: int, mlp_hidden_dim: int, target_property_dim: int, predict_atom_avg: bool = True, reference_energy_name: str = "mp-traj-d3", train_reference: bool = False, dropout: Optional[float] = None, node_aggregation: Optional[str] = None) +.. py:class:: mindchemistry.cell.orb.EnergyHead(latent_dim, num_mlp_layers, mlp_hidden_dim, target_property_dim, predict_atom_avg=True, reference_energy_name="mp-traj-d3", train_reference=False, dropout=None, node_aggregation=None) 图级能量预测头。实现用于预测分子图总能量或原子平均能量的神经网络头。支持节点级聚合、参考能量偏移和灵活的输出模式。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.GraphHead.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.GraphHead.rst index 75ae5ad7c52da0a39f5b91142a8458c68fd51323..5fd549333ca465fdf7efb061ea1b8fc1ad2dd1dd 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.GraphHead.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.GraphHead.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.GraphHead ================================= -.. py:class:: mindchemistry.cell.orb.GraphHead(latent_dim: int, num_mlp_layers: int, mlp_hidden_dim: int, target_property_dim: int, node_aggregation: str = "mean", dropout: Optional[float] = None, compute_stress: Optional[bool] = False) +.. py:class:: mindchemistry.cell.orb.GraphHead(latent_dim, num_mlp_layers, mlp_hidden_dim, target_property_dim, node_aggregation="mean", dropout=None, compute_stress=False) 图级预测头。实现可以附加到基础模型的图级预测头，用于从节点特征预测图级属性（例如应力张量），使用聚合和MLP。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.MoleculeGNS.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.MoleculeGNS.rst index c44551f329f4d6bd28dcacf08210fc60cc7a662d..0d9d62cbf93f6c8232837f275ca65b04e0ff39ad 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.MoleculeGNS.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.MoleculeGNS.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.MoleculeGNS =================================== -.. py:class:: mindchemistry.cell.orb.MoleculeGNS(num_node_in_features: int, num_node_out_features: int, num_edge_in_features: int, latent_dim: int, num_message_passing_steps: int, num_mlp_layers: int, mlp_hidden_dim: int, node_feature_names: List[str], edge_feature_names: List[str], use_embedding: bool = True, interactions: str = "simple_attention", interaction_params: Optional[Dict[str, Any]] = None) +.. py:class:: mindchemistry.cell.orb.MoleculeGNS(num_node_in_features, num_node_out_features, num_edge_in_features, latent_dim, num_message_passing_steps, num_mlp_layers, mlp_hidden_dim, node_feature_names, edge_feature_names, use_embedding=True, interactions="simple_attention", interaction_params=None) 分子图神经网络。实现用于分子性质预测的灵活模块化图神经网络，基于注意力或其他交互机制的消息传递。支持节点和边嵌入、多个消息传递步骤，以及用于复杂分子图的可定制交互层。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.NodeHead.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.NodeHead.rst index 2e422d861892a40688faea4b8b38ca0a5848bb63..7aedd3df1136f00b16f79c0b2045af96c255a73e 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.NodeHead.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.NodeHead.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.NodeHead =============================== -.. py:class:: mindchemistry.cell.orb.NodeHead(latent_dim: int, num_mlp_layers: int, mlp_hidden_dim: int, target_property_dim: int, dropout: Optional[float] = None, remove_mean: bool = True) +.. py:class:: mindchemistry.cell.orb.NodeHead(latent_dim, num_mlp_layers, mlp_hidden_dim, target_property_dim, dropout=None, remove_mean=True) 节点级预测头。 diff --git a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.Orb.rst b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.Orb.rst index fe1d53a26ce54523eb47205d9a5fa5729892f8c2..61b3bf02f83438e9dfd238b4e9d6cb6c9c59f0b5 100644 --- a/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.Orb.rst +++ b/docs/api_python/mindchemistry/cell/mindchemistry.cell.orb.Orb.rst @@ -1,7 +1,7 @@ mindchemistry.cell.orb.Orb =========================== -.. py:class:: mindchemistry.cell.orb.Orb(model: MoleculeGNS, node_head: Optional[NodeHead] = None, graph_head: Optional[GraphHead] = None, stress_head: Optional[GraphHead] = None, model_requires_grad: bool = True, cutoff_layers: Optional[int] = None) +.. py:class:: mindchemistry.cell.orb.Orb(model, node_head=None, graph_head=None, stress_head=None, model_requires_grad=True, cutoff_layers=None) Orb图回归器。将预训练的基础模型（如MoleculeGNS）与可选的节点、图和应力回归头结合，支持微调或特征提取工作流程。 diff --git a/docs/api_python/mindflow/cell/mindflow.cell.DDIMScheduler.rst b/docs/api_python/mindflow/cell/mindflow.cell.DDIMScheduler.rst index a7960e9eae2c3f7ec1cec5793cc4d2942aa4df2d..8c15c6ffc18a34206dcc1ea19b5a6d19ec4e5ea6 100644 --- a/docs/api_python/mindflow/cell/mindflow.cell.DDIMScheduler.rst +++ b/docs/api_python/mindflow/cell/mindflow.cell.DDIMScheduler.rst @@ -1,7 +1,7 @@ mindflow.cell.DDIMScheduler ============================ -.. py:class:: mindflow.cell.DDIMScheduler(num_train_timesteps: int = 1000, beta_start: float = 0.0001, beta_end: float = 0.02, beta_schedule: str = "squaredcos_cap_v2", prediction_type: str = 'epsilon', clip_sample: bool = True, clip_sample_range: float = 1.0, thresholding: bool = False, sample_max_value: float = 1.0, dynamic_thresholding_ratio: float = 0.995, rescale_betas_zero_snr: bool = False, timestep_spacing: str = "leading", compute_dtype=mstype.float32) +.. py:class:: mindflow.cell.DDIMScheduler(num_train_timesteps=1000, beta_start=0.0001, beta_end=0.02, beta_schedule="squaredcos_cap_v2", prediction_type='epsilon', clip_sample=True, clip_sample_range=1.0, thresholding=False, sample_max_value=1.0, dynamic_thresholding_ratio=0.995, rescale_betas_zero_snr=False, timestep_spacing="leading", compute_dtype=mstype.float32) `DDIMScheduler` 实现了去噪扩散概率模型DDIM中介绍的去噪过程。具体细节见 `Denoising Diffusion Implicit Models `_ 。 @@ -20,7 +20,7 @@ mindflow.cell.DDIMScheduler - **rescale_betas_zero_snr** (bool) - 是否重新缩放 betas 以使其终端 SNR 为零。这使模型能够生成非常明亮和黑暗的样本，而不是将其限制为中等亮度的样本。与 `offset_noise `_ 松散相关。默认值： ``False`` 。 - **compute_dtype** (mindspore.dtype) - 数据类型。默认值： ``mstype.float32`` ，表示 ``mindspore.float32`` 。 - .. py:method:: add_noise(original_samples: Tensor, noise: Tensor, timesteps: Tensor) + .. py:method:: add_noise(original_samples, noise, timesteps) DDIM前向加噪步骤。 @@ -42,7 +42,7 @@ mindflow.cell.DDIMScheduler 异常： - **ValueError** - 如果 `num_inference_steps` 大于 `num_train_timesteps` 。 - .. py:method:: step(model_output: Tensor, sample: Tensor, timestep: Tensor, eta: float = 0.0, use_clipped_model_output: bool = False) + .. py:method:: step(model_output, sample, timestep, eta=0.0, use_clipped_model_output=False) DDIM反向去噪步骤。 diff --git a/docs/api_python/mindflow/cell/mindflow.cell.DDPMScheduler.rst b/docs/api_python/mindflow/cell/mindflow.cell.DDPMScheduler.rst index ff7b901951a65d62e5b275e613bfa5540e401739..385d02f4007770b0f4942611079aeb53eddc57f5 100644 --- a/docs/api_python/mindflow/cell/mindflow.cell.DDPMScheduler.rst +++ b/docs/api_python/mindflow/cell/mindflow.cell.DDPMScheduler.rst @@ -1,7 +1,7 @@ mindflow.cell.DDPMScheduler ============================ -.. py:class:: mindflow.cell.DDPMScheduler(num_train_timesteps: int = 1000, beta_start: float = 0.0001, beta_end: float = 0.02, beta_schedule: str = "squaredcos_cap_v2", prediction_type: str = 'epsilon', variance_type: str = 'fixed_small_log', clip_sample: bool = True, clip_sample_range: float = 1.0, thresholding: bool = False, sample_max_value: float = 1.0, dynamic_thresholding_ratio: float=0.995, rescale_betas_zero_snr: bool = False, timestep_spacing: str = "leading", compute_dtype=mstype.float32) +.. py:class:: mindflow.cell.DDPMScheduler(num_train_timesteps=1000, beta_start=0.0001, beta_end=0.02, beta_schedule="squaredcos_cap_v2", prediction_type='epsilon', variance_type='fixed_small_log', clip_sample=True, clip_sample_range=1.0, thresholding=False, sample_max_value=1.0, dynamic_thresholding_ratio=0.995, rescale_betas_zero_snr=False, timestep_spacing="leading", compute_dtype=mstype.float32) `DDPMScheduler` 实现了去噪扩散概率模型DDPM中介绍的去噪过程。具体细节见 `Denoising Diffusion Probabilistic Models `_ 。 @@ -21,7 +21,7 @@ mindflow.cell.DDPMScheduler - **rescale_betas_zero_snr** (bool) - 是否重新缩放 betas 以使其终端 SNR 为零。这使模型能够生成非常明亮和黑暗的样本，而不是将其限制为中等亮度的样本。与 `offset_noise `_ 松散相关。默认值： ``False`` 。 - **compute_dtype** (mindspore.dtype) - 数据类型。默认值： ``mstype.float32`` ，表示 ``mindspore.float32`` 。 - .. py:method:: add_noise(original_samples: Tensor, noise: Tensor, timesteps: Tensor) + .. py:method:: add_noise(original_samples, noise, timesteps) DDPM前向加噪步骤。 diff --git a/docs/api_python/mindflow/cell/mindflow.cell.DiffusionTrainer.rst b/docs/api_python/mindflow/cell/mindflow.cell.DiffusionTrainer.rst index 4d30a2721cbf315f510e3bf3398f01eae61eed79..b945c2bd99b02198eb5e2b1825b70135a10cda9b 100644 --- a/docs/api_python/mindflow/cell/mindflow.cell.DiffusionTrainer.rst +++ b/docs/api_python/mindflow/cell/mindflow.cell.DiffusionTrainer.rst @@ -16,7 +16,7 @@ mindflow.cell.DiffusionTrainer 异常： - **TypeError** - 如果 `scheduler` 不是 `DiffusionScheduler` 类型。 - .. py:method:: get_loss(original_samples: Tensor, noise: Tensor, timesteps: Tensor, condition: Tensor = None) + .. py:method:: get_loss(original_samples, noise, timesteps, condition=None) 计算扩散过程的前向loss。 diff --git a/docs/api_python/mindflow/cell/mindflow.cell.TransformerBlock.rst b/docs/api_python/mindflow/cell/mindflow.cell.TransformerBlock.rst index 574b01e5d13fe7e8a0161fec8bd09fbd59ca6abd..4a9cd2799d66eefbc42c74f6c334e98237ef24b7 100644 --- a/docs/api_python/mindflow/cell/mindflow.cell.TransformerBlock.rst +++ b/docs/api_python/mindflow/cell/mindflow.cell.TransformerBlock.rst @@ -10,7 +10,7 @@ mindflow.cell.TransformerBlock - **num_heads** (int) - 输出的输出特征维度。 - **enable_flash_attn** (bool) - 是否使能FlashAttention。FlashAttention只支持 `Ascend` 后端。具体细节参见 `FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness `_ 。 默认值： ``False`` 。 - - **fa_dtype** (mindspore.dtype): FlashAttention计算类型。支持以下类型： `mstype.bfloat16`、 `mstype.float16`。默认值： ``mstype.bfloat16`` ，表示 ``mindspore.bfloat16`` 。 + - **fa_dtype** (mindspore.dtype) - FlashAttention计算类型。支持以下类型： `mstype.bfloat16`、 `mstype.float16`。默认值： ``mstype.bfloat16`` ，表示 ``mindspore.bfloat16`` 。 - **drop_mode** (str) - dropout方式。默认值： ``dropout`` 。支持以下类型： ``dropout`` 和 ``droppath`` 。 - **dropout_rate** (float) - dropout层丢弃的比率，在 ``[0, 1]`` 范围。默认值： ``0.0`` 。 - **compute_dtype** (mindspore.dtype) - 网络层的数据类型。默认值： ``mstype.float32`` ，表示 ``mindspore.float32`` 。 diff --git a/docs/api_python/mindflow/geometry/mindflow.geometry.Geometry.rst b/docs/api_python/mindflow/geometry/mindflow.geometry.Geometry.rst index 5646bd9740a9ebc461bc4e862fa2aec460545a7a..becaa44a906384c1c1b566fbec4dbe8130dbc757 100644 --- a/docs/api_python/mindflow/geometry/mindflow.geometry.Geometry.rst +++ b/docs/api_python/mindflow/geometry/mindflow.geometry.Geometry.rst @@ -23,7 +23,7 @@ mindflow.geometry.Geometry 异常： - **TypeError** - 如果 `name` 不是字符串。 - .. py:method:: set_sampling_config(sampling_config: SamplingConfig) + .. py:method:: set_sampling_config(sampling_config) 设置采样信息。 diff --git a/docs/api_python/mindflow/geometry/mindflow.geometry.GeometryWithTime.rst b/docs/api_python/mindflow/geometry/mindflow.geometry.GeometryWithTime.rst index 1c7a5bdf7c52c144b8b1a530d2767ff6e6e54067..4aacbf37df3fc4a48879f5a388241c1dbf065dff 100644 --- a/docs/api_python/mindflow/geometry/mindflow.geometry.GeometryWithTime.rst +++ b/docs/api_python/mindflow/geometry/mindflow.geometry.GeometryWithTime.rst @@ -34,7 +34,7 @@ mindflow.geometry.GeometryWithTime - **KeyError** - 如果 `geom_type` 为 ``"IC"``，但 `config.ic` 为 ``None``。 - **ValueError** - 如果 `geom_type` 不是 ``"BC"`` 、 ``"IC"`` 也不是 ``"domain"``。 - .. py:method:: set_sampling_config(sampling_config: SamplingConfig) + .. py:method:: set_sampling_config(sampling_config) 设置采样信息。