From f4bdc9738e330c3e40e3633bf0de06f35dc2e80c Mon Sep 17 00:00:00 2001 From: suweiyue Date: Wed, 29 Apr 2020 16:52:51 +0800 Subject: [PATCH] rm reademe --- examples/erniesage/README.md | 36 ------------------------------------ examples/erniesage/job.sh | 3 +-- 2 files changed, 1 insertion(+), 38 deletions(-) delete mode 100644 examples/erniesage/README.md diff --git a/examples/erniesage/README.md b/examples/erniesage/README.md deleted file mode 100644 index 5bc6b58..0000000 --- a/examples/erniesage/README.md +++ /dev/null @@ -1,36 +0,0 @@ -# ERNIESage in PGL - -## Introduction -In many industrial applications, there is often a special graph shown below: Text Graph. As the name implies, the node attributes of a graph consist of text, and the construction of edges provides structural information. For example, in Text Graph in search scenarios, nodes can be expressed by search terms, web page titles, and web page text, and user feedback and hyperlink information can form edge relationships. - -Text Graph - -**ERNIESage**, a model proposed by the PGL team, effectively improves the performance on Text Graph by simultaneously modeling text semantics and graph structure information. [ERNIE](https://github.com/PaddlePaddle/ERNIE) is a continual pre-training framework for language understanding launched by Baidu. It surpasses 16 tasks in Chinese and English. -When ERNIE meets Graph, it comes to the birth of ErnieSAGE(abbreviation of ERNIE SAmple aggreGatE). Its structure is shown in the figure below. The main idea is to use ERNIE as an aggregation function (Aggregators) to model the semantic and structural relationship between its own nodes and neighbor nodes. In addition, for the position-independent characteristics of neighbor nodes, Attention Mask and independent Position Embedding mechanism for neighbor blindness are designed. - -ERNIESage - -Thanks to the flexibility and usability of PGL, ERNIESage can be quickly implemented under PGL's Message Passing paradigm. - -## Dependencies -- paddlepaddle>=1.7 -- pgl>=1.1 - -## Dataformat - -## How to run - -We adopt [PaddlePaddle Fleet](https://github.com/PaddlePaddle/Fleet) as our distributed training frameworks ```config/*.yaml``` are some example config files for hyperparameters. - -```sh -# train ERNIESage in distributed gpu mode. -sh local_run.sh config/erniesage_v2_gpu.yaml - -# train ERNIESage in distributed cpu mode. -sh local_run.sh config/erniesage_v2_cpu.yaml -``` - -## Hyperparamters -- learner_type: `gpu` or `cpu`; gpu use fleet Collective mode, cpu use fleet Transpiler mode. - - diff --git a/examples/erniesage/job.sh b/examples/erniesage/job.sh index e1bc40b..8360e1c 100644 --- a/examples/erniesage/job.sh +++ b/examples/erniesage/job.sh @@ -27,13 +27,12 @@ export FLAGS_rpc_deadline=3000000 export FLAGS_rpc_retry_times=1000 if [[ $async_mode == "True" ]];then - echo 123 + echo "async_mode is True" else export FLAGS_communicator_send_queue_size=1 export FLAGS_communicator_min_send_grad_num_before_recv=0 export FLAGS_communicator_max_merge_var_num=1 # important! export FLAGS_communicator_merge_sparse_grad=0 - echo 123 fi export FLAGS_communicator_recv_wait_times=5000000 -- GitLab