MCPcopy Create free account

hub / github.com/PersiaML/PERSIA / functions

Functions720 in github.com/PersiaML/PERSIA

⨍ Functions 720◇ Types & classes 217↳ Endpoints 1
↓ 2 callersFunction_prepare_feature
( persia_training_batch: PersiaTrainingBatch, preprocess_mode: PreprocessMode = PreprocessMode.TRAIN,
persia/ctx.py:75
↓ 2 callersMethodas_mut_emb_entry_slice
(&mut self)
rust/persia-embedding-holder/src/emb_entry.rs:108
↓ 2 callersFunctionassert_ndarray_base_data
( ndarray_base_data_list: List[np.ndarray], tensors: List[torch.Tensor], use_cuda: bool, )
test/test_ctx.py:28
↓ 2 callersMethodbatched_lookup
( &self, req: Vec<(u64, usize)>, is_training: bool, )
rust/persia-embedding-server/src/embedding_parameter_service/mod.rs:162
↓ 2 callersMethodcheck_embedding_dump_done
( &self, emb_dir: &PathBuf, )
rust/persia-model-manager/src/lib.rs:178
↓ 2 callersFunctioncreate_compact_request_with_indices
()
rust/others/persia-common-benchmark/benches/serialize_inf_request.rs:258
↓ 2 callersMethodcreate_counter_vec
( &self, name: &str, help: &str, )
rust/persia-metrics/src/lib.rs:102
↓ 2 callersMethodcreate_gauge_vec
( &self, name: &str, help: &str, )
rust/persia-metrics/src/lib.rs:124
↓ 2 callersMethodcuda
(self, device_id: i32)
rust/persia-core/src/tensor.rs:272
↓ 2 callersFunctioncurrent_unix_time
()
rust/persia-incremental-update-manager/src/lib.rs:67
↓ 2 callersMethoddata_ptr
(&mut self)
rust/persia-core/src/tensor.rs:148
↓ 2 callersMethoddtype
(&self)
rust/persia-core/src/tensor.rs:306
↓ 2 callersFunctionembedding2tensor
(embedding: FeatureEmbeddingBatch, device: &Option<i32>)
rust/persia-core/src/forward.rs:333
↓ 2 callersFunctionensure_persia_service
(*args, **kwargs)
persia/helper.py:330
↓ 2 callersMethoderror_handle
( &self, err: &EmbeddingWorkerError, )
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:1320
↓ 2 callersFunctionexecute_test
(optimizer: Box<dyn Optimizable>)
rust/persia-common/src/optim.rs:342
↓ 2 callersFunctionexpect_uri
(url: url::Url)
rust/persia-common/src/message_queue.rs:27
↓ 2 callersMethodfill_elements
(&mut self, capacity: u32)
rust/persia-embedding-holder/src/array_linked_list.rs:231
↓ 2 callersFunctionfind_free_port
Check current input port is available or not. It will add the interval to input port utils the the new port is available.
persia/utils.py:83
↓ 2 callersMethodforward
( self, non_id_tensors: List[torch.Tensor], embedding_tensors: List[torch.Tensor] )
examples/src/adult-income/model.py:24
↓ 2 callersMethodforward_batched_direct
( &self, indices: IDTypeFeatureBatch, )
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:1076
↓ 2 callersFunctionforward_directly
( batch: PersiaBatchImpl, device_id: Option<i32>, )
rust/persia-core/src/forward.rs:782
↓ 2 callersMethodget
()
rust/persia-embedding-config/src/lib.rs:187
↓ 2 callersMethodget_address
(&self, replica_index: usize)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:258
↓ 2 callersMethodget_all_addresses
(&self)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:268
↓ 2 callersMethodget_configuration
( &self, )
rust/persia-embedding-server/src/embedding_parameter_service/mod.rs:150
↓ 2 callersMethodget_dtype
(&self)
rust/persia-core/src/tensor.rs:114
↓ 2 callersMethodget_emb_file_list_in_dir
( &self, dir: PathBuf, )
rust/persia-model-manager/src/lib.rs:345
↓ 2 callersMethodget_id
(&self)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:647
↓ 2 callersFunctionget_label_selector
(job_name: &str)
k8s/src/crd.rs:38
↓ 2 callersFunctionget_local_rank
Get the local rank of current process. Local rank is the rank of the process on the local machine.
persia/env.py:110
↓ 2 callersFunctionget_metrics_gateway_service_name
(job_name: &str)
k8s/src/crd.rs:34
↓ 2 callersFunctionget_nn_worker_pod_name
(job_name: &str, replica_index: usize)
k8s/src/crd.rs:22
↓ 2 callersMethodget_other_shard_dir
(&self, root_dir: &PathBuf, replica_index: usize)
rust/persia-model-manager/src/lib.rs:131
↓ 2 callersMethodget_parent_dir
(&self, root_dir: &PathBuf)
rust/persia-model-manager/src/lib.rs:138
↓ 2 callersFunctionget_rank
Get the rank of current process.
persia/env.py:104
↓ 2 callersMethodget_shard_dir
(&self, root_dir: &PathBuf)
rust/persia-model-manager/src/lib.rs:124
↓ 2 callersFunctionindices_add_prefix
(indices: &mut IDTypeFeatureBatch, config: &EmbeddingConfig)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:403
↓ 2 callersFunctionindices_to_hashstack_indices
( indices: &mut IDTypeFeatureBatch, config: &EmbeddingConfig, )
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:348
↓ 2 callersMethodinit_master_discovery_service
(&mut self, master_addr: Option<String>)
rust/persia-core/src/lib.rs:228
↓ 2 callersMethodinit_rpc_client_with_addr
(&self, addr: String)
rust/persia-core/src/lib.rs:184
↓ 2 callersMethodinsert_between
(&mut self, prev_index: u32, next_index: u32, value: T)
rust/persia-embedding-holder/src/array_linked_list.rs:330
↓ 2 callersMethodis_empty
Checks if the list is empty.
rust/persia-embedding-holder/src/array_linked_list.rs:791
↓ 2 callersMethodis_master
(&self)
rust/persia-embedding-config/src/lib.rs:205
↓ 2 callersMethodlaunch
(&mut self, num_workers: usize)
rust/persia-core/src/forward.rs:530
↓ 2 callersMethodload_array_linked_list
( &self, file_path: PathBuf, )
rust/persia-model-manager/src/lib.rs:276
↓ 2 callersMethodload_torch_state_dict
Load a Pytorch state dict from the source directory and apply to `torch_instance`. Arguments: torch_instance (torch.nn.Module or
persia/ctx.py:584
↓ 2 callersMethodlookup_batched_all_slots
( &self, indices: &mut IDTypeFeatureBatch, requires_grad: bool, )
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:874
↓ 2 callersMethodmark_embedding_dump_done
( &self, emb_dir: PathBuf, num_internal_shards: usize, )
rust/persia-model-manager/src/lib.rs:156
↓ 2 callersMethodmodel_manager_status
(&self)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:225
↓ 2 callersFunctionndarray_f32_to_f16
(input: &Array2<f32>)
rust/persia-common/src/lib.rs:157
↓ 2 callersMethodparent
(&self)
rust/persia-storage/src/lib.rs:76
↓ 2 callersMethodpop_from_buffer
(&mut self)
rust/persia-core/src/forward.rs:459
↓ 2 callersMethodprepare_features
r"""This function converts data from ``PersiaTrainingBatch`` to ``torch.Tensor``. :class:`PersiaTrainingBatch` contains non_id_type_features,
persia/ctx.py:450
↓ 2 callersFunctionprocess
(df_dataset: pd.core.frame.DataFrame, filepath: str)
examples/src/adult-income/data/data_preprocess.py:11
↓ 2 callersMethodprocess_status
(&self, status: Vec<EmbeddingModelManagerStatus>)
rust/persia-core/src/rpc.rs:243
↓ 2 callersMethodraw_data_ptr
(&mut self)
rust/persia-core/src/tensor.rs:294
↓ 2 callersMethodread_to_end_speedy
(&self)
rust/persia-storage/src/lib.rs:116
↓ 2 callersMethodrecord
(&self, stream: CudaStreamPtr)
rust/persia-core/src/cuda/cuda_event_pool.rs:27
↓ 2 callersFunctionreload_env
Reload the environment.
persia/env.py:78
↓ 2 callersMethodreq_type
(&self)
rust/others/persia-nats-marcos/src/lib.rs:28
↓ 2 callersFunctionserialize_tear_indices
( meta: &CompactBatchEmbeddingsRequestWithoutIndices, indices: &Vec<u64>, )
rust/others/persia-common-benchmark/benches/serialize_inf_request.rs:291
↓ 2 callersMethodset
( self, replica_size: Optional[int] = None, replica_index: Optional[int] = None,
persia/env.py:58
↓ 2 callersFunctionsetup
(sphinx)
docs/conf.py:155
↓ 2 callersMethodstate_initialization
(&self, state: &mut [f32], dim: usize)
rust/persia-common/src/optim.rs:299
↓ 2 callersMethodsubject_ident
(&self)
rust/others/persia-nats-marcos/src/lib.rs:16
↓ 2 callersMethodsubscribe_subject_ident
(&self)
rust/others/persia-nats-marcos/src/lib.rs:20
↓ 2 callersMethodto_bytes
Serialize the [`PersiaBatchImpl`] to bytes.
rust/persia-core/src/data.rs:256
↓ 2 callersMethodwait_for_serving
(&self)
rust/persia-core/src/lib.rs:314
↓ 2 callersMethodwrite_all
(&self, content: Vec<u8>)
rust/persia-storage/src/lib.rs:124
↓ 2 callersMethodwrite_all_speedy
(&self, content: &W)
rust/persia-storage/src/lib.rs:138
↓ 1 callersMethod__init__
Arguments: initialization_method (str): the PyTorch distributed initialization_method method, support tcp and fil
persia/distributed.py:98
↓ 1 callersMethod__init__
( self, dataset: IterableDatasetBase, forward_buffer_size: int = 10, timeout_m
persia/data.py:228
↓ 1 callersFunction_check_finite
Check if all tensors in the input list contain only finite elements. Arguments: tensors (List[torch.Tensor]): list of tensor to be checke
persia/ctx.py:28
↓ 1 callersMethod_get_master_addr
Get leader(rank 0) ip address.
persia/ctx.py:865
↓ 1 callersMethod_init_embedding_worker_rpc_client
Initialize the embedding worker rpc clients.
persia/ctx.py:877
↓ 1 callersFunction_is_port_available
(port: int)
persia/utils.py:69
↓ 1 callersFunction_launch_data_loader
(replica_num: int, env: dict)
persia/helper.py:73
↓ 1 callersFunction_launch_nats_server
(port: int)
persia/helper.py:113
↓ 1 callersFunction_launch_nn_worker
(nproc_per_node: int, env: dict)
persia/helper.py:91
↓ 1 callersFunction_ndarray_check
r"""Check if the dtype, shape and batch_size is valid or not. Arguments: data (np.ndarray): data that needs to be checked. data_n
persia/embedding/data.py:39
↓ 1 callersMethod_on_backward
Update the embeddings' gradients Arguments: loss_scale (float): the loss that scaled by GradScalar, loss_scale always equal
persia/ctx.py:926
↓ 1 callersMethod_prepare
Do some preparation to init `DataCtx`.
persia/ctx.py:329
↓ 1 callersFunction_select_bagua_algorithm
Select corresponding bagua algorithm for current training. Arguments: algorithm (str): name of Bagua algorithm. model (torch.nn.M
persia/distributed.py:204
↓ 1 callersFunction_set_default_logger
r"""Set the default logger. Arguments: name (str): default logger name. logging.Logger
persia/logger.py:105
↓ 1 callersFunctionadam_avx2
( adam_m: &mut [f32], adam_v: &mut [f32], beta1_power: f32, beta2_power: f32, emb: &mut [f
rust/persia-simd/src/lib.rs:147
↓ 1 callersFunctionadd
(client: Client, name: &str, namespace: &str)
k8s/src/finalizer.rs:6
↓ 1 callersMethodadd_gradient
( &mut self, slot_name: String, data_ptr: u64, shape: [usize; 2], is_f
rust/persia-core/src/backward.rs:86
↓ 1 callersMethodadd_id_type_feature
Add id_type_feature which is a sparse matrix into [`PersiaBatchImpl`] with required name.
rust/persia-core/src/data.rs:183
↓ 1 callersMethodadd_id_type_feature_with_single_id
Add id_type_feature which is a vector into [`PersiaBatchImpl`] with required name.
rust/persia-core/src/data.rs:208
↓ 1 callersMethodadd_label
Add label into [`PersiaBatchImpl`] with optional name.
rust/persia-core/src/data.rs:171
↓ 1 callersMethodadd_meta
Add binary data into [`PersiaBatchImpl`].
rust/persia-core/src/data.rs:251
↓ 1 callersMethodadd_non_id_type_feature
Add non_id_type_feature into [`PersiaBatchImpl`] with optional name.
rust/persia-core/src/data.rs:159
↓ 1 callersMethodadd_skipped_gradient
(&mut self, slot_name: String)
rust/persia-core/src/backward.rs:76
↓ 1 callersMethodapply
Register sparse optimizer to embedding server.
persia/embedding/optim.py:13
↓ 1 callersMethodas_emb_entry_slice
(&self)
rust/persia-embedding-holder/src/emb_entry.rs:112
↓ 1 callersMethodbackward
Update the parameters of the current dense model and embedding model. Arguments: loss (torch.Tensor): loss of current batch.
persia/ctx.py:893
↓ 1 callersMethodbuffer_input_thread
(&self)
rust/persia-incremental-update-manager/src/lib.rs:251
↓ 1 callersMethodbuffer_output_thread
(&self)
rust/persia-incremental-update-manager/src/lib.rs:272
↓ 1 callersMethodcan_forward_batched
(&self, batcher_idx: usize)
rust/persia-embedding-server/src/embedding_worker_service/mod.rs:991
← previousnext →101–200 of 720, ranked by callers