DeepRec
DeepRec copied to clipboard
DeepRec is a high-performance recommendation deep learning framework based on TensorFlow. It is hosted in incubation in LF AI & Data Foundation.
1. build DeepRec with image: registry.cn-shanghai.aliyuncs.com/pai-dlc-share/deeprec-developer:deeprec-dev-cpu-py36-ubuntu18.04(https://github.com/alibaba/DeepRec, 'Prepare for installation') 2. enabling auto-micro-batch (file /modelzoo/WDL/train.py)  3. then run 'python train.py' and it crashes:  4. if setting micro_batch_num = 1,...
``` from tensorflow.python import pywrap_tensorflow reader = pywrap_tensorflow.NewCheckpointReader(latest_checkpoint) var_to_shape_map = reader.get_variable_to_shape_map() for key in var_to_shape_map: print(reader.get_tensor(key)) ``` I want export the value of embedding variables, and I test it in...
We first test the **star_server** protocol on the CPU machine, and the training task runs normally. Now, we want to switch to the GPU machine. The cluster info is 2...
When I was reading https://github.com/alibaba/DeepRec/blob/main/triton/tensorflow_backend_tf.cc#L941 https://github.com/alibaba/DeepRec/blob/main/triton/tensorflow_backend_tf.cc#L932 I wonder where function clear_allocator_type() in line 941 and set_allocator_type() in line 932 are defined. I did not find any file in tensorflow related...
This PR is to enable SOK lookup support DeepRec EmbeddingVariableGPU. It includes 4 parts: 1. move EmbeddingVarGPU from kernel to framework 2. add hugectr as a submodule under SOK 3....
contains these PRs: * https://github.com/alibaba/DeepRec/pull/435 * https://github.com/alibaba/DeepRec/pull/440 * https://github.com/alibaba/DeepRec/pull/442 * https://github.com/alibaba/DeepRec/pull/443 And there are still many warnings when use GCC11 or GCC12, needs to remove `-Werror`