DeepRec-AI/HybridBackend issues and pull requests

#160 - Support default values for filename dataset

Pull Request - State: open - Opened by 2sin18 5 months ago

#159 - Op type not registered 'HbGetNcclId' in binary

Issue - State: open - Opened by ZhuYuJin 6 months ago

#158 - fix pyarrow error

Pull Request - State: open - Opened by Nov11 7 months ago

#157 - Support default value for fields not exist

Pull Request - State: closed - Opened by 2sin18 7 months ago
Labels: enhancement

#156 - ParquetDataset support configuration with default value

Issue - State: open - Opened by Markz2z 8 months ago
Labels: enhancement

#155 - Training is very slow

Issue - State: closed - Opened by dixingxing0 12 months ago - 4 comments

#154 - Error in multi-card in a single machine mode

Issue - State: closed - Opened by dixingxing0 12 months ago

#153 - Train got error died with <Signals.SIGSEGV: 11>

Issue - State: closed - Opened by dixingxing0 12 months ago - 3 comments

#152 - hb.data.ParquetDataset in hb.estimator.train_and_evaluate will loss data

Issue - State: closed - Opened by karterotte 12 months ago - 1 comment

#151 - No OpKernel was registered to support Op 'HbSparseSegmentMeanGrad1' used by node

Issue - State: open - Opened by karterotte about 1 year ago

#150 - [DOC] Add user docs for data deduplication and so forth.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#149 - [DATA] Implement zero-copied string dtype and accelerate shuffle.

Pull Request - State: closed - Opened by francktcheng over 1 year ago - 1 comment

#148 - [DATA] Implement zero-copied string dtype and accelerate shuffle.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#147 - Exception occurs when call `batch` with ragged tensor

Issue - State: closed - Opened by DelightRun over 1 year ago

#146 - [DATA] Implement zero-copied string dtype and accelerate shuffle.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#145 - [DATA] Implement zero-copied string dtype and accelerate shuffle.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#144 - [DATA] Implement zero-copied string dtype and accelerate shuffle.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#143 - [DIST] Set data_sync_drop_remainder as true by default.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#142 - [DIST] Set data_sync_drop_remainder as true by default.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#141 - [CI] Upgrade DeepRec docker to 2302

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#140 - [DIST] Set data_sync_drop_remainder as true by default.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#139 - [CI] Add more build information

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#138 - Throughput is lower than TFRecords when there are many strings in Parquets file

Issue - State: open - Opened by deepllz over 1 year ago

#137 - [DIST] Rewrite NanTensorHook for SyncReplicasDataset.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#136 - [CI] Refines logging and cibuild

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#135 - [DATA] Refine data deduplication and add example.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#134 - [DIST] Refines NCCL logging messages for debugging

Pull Request - State: closed - Opened by 2sin18 over 1 year ago

#133 - [DIST] Using empty data at the end of SyncReplicasDataset.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#132 - [CI] Upgrade to 1.0.0

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#131 - [EMB] Add experimental support for embedding service acceleration

Pull Request - State: closed - Opened by 2sin18 over 1 year ago

#130 - [DIST] Implement a hierarchical embedding lookup.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#129 - [DATA] Support transfer optimization

Pull Request - State: closed - Opened by 2sin18 over 1 year ago

#128 - [DATA] Support ORC format and data deduplication

Pull Request - State: closed - Opened by 2sin18 over 1 year ago
Labels: enhancement

#127 - [DIST] Use variable_scope with partitioner for sharded deeprecev.

Pull Request - State: closed - Opened by francktcheng over 1 year ago - 1 comment

#126 - [DIST] Refine naming of deeprecev variables in unsharded cases.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#125 - Deeprec hangs in distributed mode.

Issue - State: open - Opened by silingtong123 over 1 year ago

#124 - [DIST] Support a standalone evaluation in `hb.keras`.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#123 - [DIST] Fix: Shared embedding shape rewriting.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#122 - Failed to train with multiple GPUs in single node

Issue - State: closed - Opened by ZhuYuJin over 1 year ago

#121 - hb.data.ParquetDataset will discard some data

Issue - State: closed - Opened by silingtong123 over 1 year ago

#120 - init_from_checkpoint throw Exception when using hb.keras.Model

Issue - State: closed - Opened by karterotte over 1 year ago - 1 comment

#119 - [CI] Upgrade DeepRec

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#118 - hb.keras.model evaluate error

Issue - State: closed - Opened by karterotte over 1 year ago

#117 - EmbeddingLookupRewritingForDeepRecEV Add "part0" to op-name twice

Issue - State: closed - Opened by karterotte over 1 year ago

#116 - [CI] Fix GitHub workflow files

Pull Request - State: closed - Opened by 2sin18 over 1 year ago

#115 - [DIST] Support standalone evaluation and prediction in `hb.Estimator`.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#114 - [DIST] Support pipeline-based semi-synchronous training.

Pull Request - State: closed - Opened by francktcheng over 1 year ago

#113 - [DIST] Refactor directories of embedding sharding

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#112 - [DATA] Refines data transfer prefetching and synchronization

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment
Labels: enhancement

#111 - [DIST] Support Non-intrusive embedding APIs

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment
Labels: enhancement

#110 - [DATA] Introduce new RebatchDataset to replace rebatch and rectify

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment
Labels: bug

#109 - [DOC] Refines documentation for new tabular dataset API

Pull Request - State: closed - Opened by 2sin18 over 1 year ago

#108 - [DATA] Introduce new tabular dataset API with row-wise shuffling

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment
Labels: enhancement

#107 - Row-wise shuffling required

Issue - State: closed - Opened by 2sin18 over 1 year ago
Labels: enhancement

#106 - QR code is invalid

Issue - State: closed - Opened by co63oc over 1 year ago - 2 comments

#105 - [DATA] Support skipping corrupted data

Pull Request - State: closed - Opened by 2sin18 over 1 year ago
Labels: enhancement

#104 - ParquetDataset should be able to skip corrupted data

Issue - State: closed - Opened by 2sin18 over 1 year ago
Labels: enhancement

#103 - [CI] Upgrades third-party libraries

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment

#102 - How to place the embeddings on gpu?

Issue - State: closed - Opened by taoyun951753 over 1 year ago

#101 - Following the BUILD.md tutorial, something is wrong

Issue - State: closed - Opened by taoyun951753 over 1 year ago

#100 - merge embedding table

Issue - State: closed - Opened by zhaozheng09 over 1 year ago

#99 - How to pad a column to specific size when using hb.data.ParquetDataset ?

Issue - State: closed - Opened by taoyun951753 over 1 year ago

#98 - [DIST] Support sync training of imbalanced data across devices

Pull Request - State: closed - Opened by 2sin18 over 1 year ago - 1 comment
Labels: enhancement

#97 - ParquetDataset benchmark add tfrecord data

Issue - State: closed - Opened by welsonzhang over 1 year ago

#96 - Sync training with ParquetDataset, Use PS-Worker，The system may block because some worker stop early.

Issue - State: closed - Opened by zhbhhb over 1 year ago

#95 - DLRM model on A100 8cards training

Issue - State: closed - Opened by zhaozheng09 almost 2 years ago - 1 comment

#94 - How to realize gradient truncation function in HB Pkg

Issue - State: closed - Opened by taoyun951753 almost 2 years ago

#93 - tf.keras.layers.DenseFeatures api as the candidate of hb.feature_column.DenseFeatures can not work with tf.feature_column.shared_embedding_columns

Issue - State: closed - Opened by taoyun951753 almost 2 years ago - 1 comment

#92 - Multi worker jobs run error under hb.keras api

Issue - State: closed - Opened by taoyun951753 almost 2 years ago

#91 - [DIST] Multiple enhancement in `hb.keras`.

Pull Request - State: closed - Opened by francktcheng almost 2 years ago

#90 - Metrics in HB multi outputs job don't get correctly

Issue - State: closed - Opened by taoyun951753 almost 2 years ago

#89 - feature_column bucket_size is 6, use 8 gpus, then worker-5 and worker-6 'save/RestoreV2' failed

Issue - State: open - Opened by zhbhhb almost 2 years ago

#88 - the EarlyStopping callback not working well on multi worker distribute training job

Issue - State: open - Opened by taoyun951753 almost 2 years ago

#87 - model.summary didn't show model layers

Issue - State: closed - Opened by karterotte almost 2 years ago
Labels: bug

#86 - hb.keras.Model.load_weight support logging details

Issue - State: closed - Opened by karterotte almost 2 years ago

#85 - hb.keras.Model's fit() func support dataset multiple labels

Issue - State: closed - Opened by karterotte almost 2 years ago - 1 comment

#84 - Dataset iterator can't be warpped in the hybridBackend scope

Issue - State: open - Opened by fuhailin almost 2 years ago

#83 - [DIST] Support loading weights and Functional API in Keras.

Pull Request - State: closed - Opened by francktcheng almost 2 years ago

#82 - Support keras.Model convert to hb.keras.Model by one function

Issue - State: closed - Opened by karterotte almost 2 years ago

#81 - Support Keras load_weights to do model transfer or inherit

Issue - State: closed - Opened by karterotte almost 2 years ago

#80 - [CI] Refactors directory layout

Pull Request - State: closed - Opened by 2sin18 almost 2 years ago - 1 comment

#79 - [DIST] Support data sync and evaluation using embedding_lookup_* API

Pull Request - State: closed - Opened by 2sin18 almost 2 years ago - 2 comments

#78 - error: Variables not initialized: communicator/1/HbNcclCommHandleOp

Issue - State: closed - Opened by shijieliu almost 2 years ago

#77 - [DIST] Support hybrid parallelism for embedding_lookup_* API

Pull Request - State: closed - Opened by 2sin18 about 2 years ago - 2 comments
Labels: enhancement

#76 - support keras fit history in estimator's train_and_evaluate

Issue - State: closed - Opened by karterotte about 2 years ago - 1 comment
Labels: enhancement

#75 - [DATA] Supports `hb.data.Iterator`

Pull Request - State: closed - Opened by 2sin18 about 2 years ago - 1 comment
Labels: enhancement

#74 - Feature Request: Supports prefetching data to GPU

Issue - State: closed - Opened by 2sin18 about 2 years ago - 3 comments
Labels: enhancement

#73 - [CI] Fixes GitHub workflows

Pull Request - State: closed - Opened by 2sin18 about 2 years ago - 1 comment

#72 - [CI] Refactors tools and workflows

Pull Request - State: closed - Opened by 2sin18 about 2 years ago - 1 comment

#71 - [Data] Fix nested list support

Pull Request - State: closed - Opened by 2sin18 about 2 years ago - 1 comment
Labels: bug

#70 - [Data] Fix support for nested list

Pull Request - State: closed - Opened by 2sin18 about 2 years ago
Labels: bug

#69 - to_sparse failed for Value with ragged_rank > 1 read from parquet file

Issue - State: open - Opened by SamJia about 2 years ago - 9 comments
Labels: bug

#68 - [DOC] Refine README

Pull Request - State: closed - Opened by 2sin18 about 2 years ago

#67 - support ARROW_NUM_THREADS in ParquetDataset

Issue - State: closed - Opened by karterotte about 2 years ago - 1 comment

#66 - [TRAIN] Support hb.train.Policy for easier training customization

Pull Request - State: closed - Opened by 2sin18 over 2 years ago - 1 comment

#65 - [EMB] Refine embedding lookup and saved model export

Pull Request - State: closed - Opened by 2sin18 over 2 years ago - 1 comment

#64 - [CI] Improve modularity and compatibility

Pull Request - State: closed - Opened by 2sin18 over 2 years ago - 1 comment
Labels: enhancement

#63 - [CI] Sort imports and refine wheel building

Pull Request - State: closed - Opened by 2sin18 over 2 years ago - 1 comment

#62 - hybridbackend 0.6.0a2 version raise ValueError when ParquetDataset wrapped by parallel_interleave ops

Issue - State: closed - Opened by fuhailin over 2 years ago

#61 - [CI] Upgrade version to 0.5.4

Pull Request - State: closed - Opened by 2sin18 over 2 years ago

GitHub / DeepRec-AI/HybridBackend issues and pull requests