search

NVIDIA-Merlin / HugeCTR

HugeCTR is a high efficiency GPU framework designed for Click-Through-Rate (CTR) estimating training
Apache License 2.0
905 stars 196 forks source link

[BUG]preprocess.sh 1 criteo failed with 'Schema' object has no attribute 'write' #435

Open SeekPoint opened 6 months ago

SeekPoint commented 6 months ago

root@0da37f00b33c:/share/criteo_data# bash preprocess.sh 1 criteo_tmp nvt 0 0 0 Warning: existing criteo_tmp is erased Preprocessing script: NVTabular Getting the first few examples from the uncompressed dataset... Counting the number of samples in day_1 dataset... The first 45840617 examples will be used in day_1 dataset. Shuffling dataset... Preprocessing... Splitting into 36672493-sample training, 4584062-sample val, and 4584062-sample test datasets... /usr/local/lib/python3.10/dist-packages/merlin/dtypes/mappings/tf.py:52: UserWarning: Tensorflow dtype mappings did not load successfully due to an error: No module named 'tensorflow' warn(f"Tensorflow dtype mappings did not load successfully due to an error: {exc.msg}") /usr/local/lib/python3.10/dist-packages/merlin/dtypes/mappings/torch.py:43: UserWarning: PyTorch dtype mappings did not load successfully due to an error: No module named 'torch' warn(f"PyTorch dtype mappings did not load successfully due to an error: {exc.msg}") 2023-12-08 17:36:24,676 NVTabular processing 2023-12-08 17:36:24,699 To route to workers diagnostics web server please install jupyter-server-proxy: python -m pip install jupyter-server-proxy 2023-12-08 17:36:24,702 State start 2023-12-08 17:36:24,712 Scheduler at: tcp://127.0.0.1:40579 2023-12-08 17:36:24,712 dashboard at: 127.0.0.1:8787 2023-12-08 17:36:24,781 Start Nanny at: 'tcp://127.0.0.1:40985' /usr/local/lib/python3.10/dist-packages/merlin/dtypes/mappings/tf.py:52: UserWarning: Tensorflow dtype mappings did not load successfully due to an error: No module named 'tensorflow' warn(f"Tensorflow dtype mappings did not load successfully due to an error: {exc.msg}") /usr/local/lib/python3.10/dist-packages/merlin/dtypes/mappings/torch.py:43: UserWarning: PyTorch dtype mappings did not load successfully due to an error: No module named 'torch' warn(f"PyTorch dtype mappings did not load successfully due to an error: {exc.msg}") 2023-12-08 17:36:26,969 - distributed.preloading - INFO - Creating preload: dask_cuda.initialize 2023-12-08 17:36:26,969 Creating preload: dask_cuda.initialize 2023-12-08 17:36:26,969 - distributed.preloading - INFO - Import preload module: dask_cuda.initialize 2023-12-08 17:36:26,969 Import preload module: dask_cuda.initialize 2023-12-08 17:36:26,993 Run preload setup: dask_cuda.initialize 2023-12-08 17:36:27,086 Start worker at: tcp://127.0.0.1:46779 2023-12-08 17:36:27,086 Listening to: tcp://127.0.0.1:46779 2023-12-08 17:36:27,086 Worker name: 0 2023-12-08 17:36:27,086 dashboard at: 127.0.0.1:38661 2023-12-08 17:36:27,086 Waiting to connect to: tcp://127.0.0.1:40579 2023-12-08 17:36:27,086 ------------------------------------------------- 2023-12-08 17:36:27,086 Threads: 1 2023-12-08 17:36:27,086 Memory: 251.56 GiB 2023-12-08 17:36:27,087 Local Directory: /tmp/dask-worker-space/worker-1fd2e7px 2023-12-08 17:36:27,087 Starting Worker plugin PreImport-047eaac1-e793-4dd5-b59e-bd5028f4f909 2023-12-08 17:36:27,087 Starting Worker plugin RMMSetup-c63a5516-1d21-4b3e-8187-a1f883e8fb5e 2023-12-08 17:36:27,087 Starting Worker plugin CPUAffinity-1188f5ac-00e8-4229-9bea-18511f88e776 2023-12-08 17:36:27,087 ------------------------------------------------- 2023-12-08 17:36:27,097 Register worker <WorkerState 'tcp://127.0.0.1:46779', name: 0, status: init, memory: 0, processing: 0> 2023-12-08 17:36:27,103 Starting worker compute stream, tcp://127.0.0.1:46779 2023-12-08 17:36:27,103 Starting established connection to tcp://127.0.0.1:48730 2023-12-08 17:36:27,103 Registered to: tcp://127.0.0.1:40579 2023-12-08 17:36:27,103 ------------------------------------------------- 2023-12-08 17:36:27,104 Starting established connection to tcp://127.0.0.1:40579 2023-12-08 17:36:27,142 Receive client connection: Client-508fa7e6-95f0-11ee-8237-0242ac110004 2023-12-08 17:36:27,143 Starting established connection to tcp://127.0.0.1:48732 2023-12-08 17:36:27,147 Run out-of-band function 'reinitialize' /usr/local/lib/python3.10/dist-packages/merlin/core/utils.py:361: FutureWarning: The client argument is deprecated from DaskExecutor and will be removed in a future version of NVTabular. By default, a global client in the same python context will be detected automatically, and merlin.utils.set_dask_client (as well as Distributed and Serial) can be used for explicit control. warnings.warn( 2023-12-08 17:36:45,511 Preprocessing 2023-12-08 17:36:45,863 Train Datasets Preprocessing..... 2023-12-08 17:36:46,095 Run out-of-band function 'clean_worker_cache' 2023-12-08 17:36:46,332 Run out-of-band function 'clean_worker_cache' 2023-12-08 17:37:07,465 Run out-of-band function 'clean_worker_cache' 2023-12-08 17:37:08,855 Run out-of-band function 'clean_worker_cache' Traceback (most recent call last): File "/share/criteo_data/criteo_script/preprocess_nvt.py", line 418, in process_NVT(args) File "/share/criteo_data/criteo_script/preprocess_nvt.py", line 216, in process_NVT workflow.transform(train_ds_iterator).to_hugectr( File "/usr/local/lib/python3.10/dist-packages/merlin/io/dataset.py", line 1070, in to_hugectr self.schema.write(output_path) AttributeError: 'Schema' object has no attribute 'write'

jershi425 commented 6 months ago

Hi @SeekPoint , thanks for the feedback. We will fix this issue and let you know.