GithubHelp home page GithubHelp logo

paddlepaddle / research Goto Github PK

View Code? Open in Web Editor NEW
1.7K 46.0 793.0 295.51 MB

novel deep learning research works with PaddlePaddle

License: Apache License 2.0

Python 58.04% Shell 3.42% C++ 0.59% Cuda 0.14% Jupyter Notebook 35.88% Makefile 0.01% Jsonnet 0.26% Perl 1.66% C 0.01%
deep-learning computer-vision nlp knowledge-graph spatial-temporal data-mining

research's Introduction

Research

发布基于飞桨的前沿研究工作,包括CV、NLP、KG、STDM等领域的顶会论文和比赛冠军模型。

目录

计算机视觉

任务类型 目录 简介 论文链接
图像检索 GNN-Re-Ranking 基于GNN的快速图像检索Re-Ranking。 https://arxiv.org/abs/2012.07620v2
车流统计 VehicleCounting AICITY2020 车流统计竞赛datasetA TOP1 方案。 -
车辆再识别 PaddleReid 给定目标车辆,在检索库中检索同id车辆,支持多种特征子网络。 -
车辆异常检测 AICity2020-Anomaly-Detection 在监控视频中检测车辆异常情况,例如车辆碰撞、失速等。 -
医学图像分析 AGEchallenge 任务:在AS-OCT图像的公共数据集上进行闭角型分类和巩膜突点定位;基线模型:对应以上各任务的基线模型。 -
光流估计 PWCNet 基于金字塔式处理,逐层学习细部光流,设计代价容量函数三原则的CNN模型,用于光流估计。 https://arxiv.org/abs/1709.02371
语义分割 SemSegPaddle 针对多个数据集的图像语义分割模型的实现,包括Cityscapes、Pascal Context和ADE20K。 -
轻量化检测 astar2019 百度之星轻量化检测比赛评测工具。 -
地标检索与识别 landmark 基于检索的地标检索与识别系统,支持地标型与非地标型识别、识别与检索结果相结合的多重识别结果投票和重新排序。 https://arxiv.org/abs/1906.03990
图像分类 webvision2018 模型利用重加权网络(URNet)缓解web数据中偏倚和噪声的影响,进行web图像分类。 https://arxiv.org/abs/1811.00700
图像分类 CLPI 模型利用一个Lesion Generator改善了糖尿病视网膜病变图像分级的模型性能,理论上可用于所有希望实现局部+整体模型分析的场景 -
图像分类 [RSNA-IHD](CV/Effective Transformer-based Solution for RSNA Intracranial Hemorrhage Detection) 提出了一种有效的颅内出血检测(IHD)方法,其性能超过了在RSNA-IHD竞赛(2019)中获胜的解决方案。与此同时,与获胜者的解决方案相比,我们的模型只有其20%的参数量和10%的FLOPs https://arxiv.org/abs/2205.07556
小样本学习 PaddleFSL 小样本学习工具包,可复现多个常用基线方法在多个图片分类数据集上的汇报效果 -
迁移学习 SMILE 提出了一种自蒸馏样本混合迁移学习框架,适用于小样本图片分类 https://arxiv.org/abs/2103.13941

自然语言处理

任务类型 目录 简介 论文链接
中文词法分析 LAC(Lexical Analysis of Chinese) 百度自主研发中文特色模型词法分析任务,集成了中文分词、词性标注和命名实体识别任务。输入是一个字符串,而输出是句子中的词边界和词性、实体类别。 -
主动对话 DuConv 机器根据给定知识信息主动引领对话进程完成设定的对话目标。 https://www.aclweb.org/anthology/P19-1369/
语义解析 Text2SQL-BASELINE 输入自然语言问题和相应的数据库,生成与问题对应的 SQL 查询语句,通过执行该 SQL 可得到问题的答案。 -
多轮对话 DAM 开放领域多轮对话匹配的深度注意力机制模型,根据多轮对话历史和候选回复内容,排序出最合适的回复。 http://aclweb.org/anthology/P18-1103
阅读理解 DuReader 数据集:大规模、面向真实应用、由人类生成的中文阅读理解数据集,聚焦于真实世界中的不限定领域的问答任务;基线系统:针对DuReader数据集实现的经典BiDAF模型。 https://www.aclweb.org/anthology/W18-2605/
关系抽取 ARNOR 数据集:用于对远程监督关系提取模型进行句子级别的评价;模型:基于注意力正则化识别噪声数据,通过bootstrap方法逐步选择出高质量的标注数据。 https://www.aclweb.org/anthology/P19-1135/
机器翻译 JEMT 模型的输入端包括文字信息及发音信息,嵌入层融合文字信息和发音信息进行翻译。 https://arxiv.org/abs/1810.06729
阅读理解 KTNET 模型将知识库中的知识整合到预先训练好的上下文表示中,利用丰富的知识增强机器阅读理解的预训练语言表示。 https://www.aclweb.org/anthology/P19-1226
对话生成 PLATO 基于隐空间的端到端的预训练对话生成模型,可以灵活支持多种对话,包括闲聊、知识聊天、对话问答等。 http://arxiv.org/abs/1910.07931
阅读理解 DuReader-Robust-BASELINE 数据集:DuReader-robust,中文数据集,用于全面评价机器阅读理解模型的鲁棒性;基线系统:针对该数据集,基于ERNIE实现的阅读理解基线系统。 https://arxiv.org/abs/2004.11142
对话生成 AKGCM 包含知识增强图、知识选择和知识感知响应生成器的聊天机器人。 https://www.aclweb.org/anthology/D19-1187/
机器翻译 MAL 多智能体端到端联合学习框架,通过多个智能体的互相学习提升翻译质量。 https://arxiv.org/abs/1909.01101
对话生成 MMPMS 针对开放域对话中一对多问题,利用多映射机制和后验映射选择模块进行多样性、丰富化的对话生成。 https://arxiv.org/abs/1906.01781
阅读理解 MRQA2019-BASELINE 机器阅读理解任务的基线模型,基于ERNIE预训练模型,支持多GPU微调预测。 -
阅读理解 D-NET 预训练及微调框架,包含多任务学习及多预训练模型的融合,用于阅读理解模型的生成。 https://www.aclweb.org/anthology/D19-5828/
建议挖掘 MPM 利用多视角架构来学习表示和双向transformer编码器进行论坛评论建议挖掘。 https://www.aclweb.org/anthology/S19-2216/
多文档摘要 ACL2020-GraphSum 基于图表示的生成式多文档摘要模型,将显式图结构信息引入到端到端摘要生成过程中。 https://www.aclweb.org/anthology/2020.acl-main.555.pdf
融合多种对话类型的对话式推荐 ACL2020-DuRecDial 提出新任务:融合闲聊、任务型对话、问答和推荐等多种对话类型的对话式推荐,构建DuRecDial数据集,提出具有多对话目标驱动策略机制的对话生成框架。 https://www.aclweb.org/anthology/2020.acl-main.98/
面向推荐的对话 Conversational-Recommendation-BASELINE 融合人机对话系统和个性化推荐系统,定义新一代智能推荐技术,该系统先通过问答或闲聊收集用户兴趣和偏好,然后主动给用户推荐其感兴趣的内容,比如餐厅、美食、电影、新闻等。 -
稠密段落检索 ACL2021-PAIR 基于以段落相似度为中心的相似度关系提升稠密段落检索,基于知识蒸馏进行采样,采用两阶段训练方式。 https://aclanthology.org/2021.findings-acl.191/
任务式对话 EMNLP2022-Q-TOD 自然语言查询驱动的任务式对话系统,提出由查询生成、知识检索和回复生成组成的三阶段新框架。 https://arxiv.org/abs/2210.07564

知识图谱

任务类型 目录 简介 论文链接
知识图谱表示学习 CoKE 百度自主研发语境化知识图谱表示学习框架CoKE,在知识图谱链接预测和多步查询任务上取得学界领先效果。 https://arxiv.org/abs/1911.02168
关系抽取 DuIE_Baseline 语言与智能技术竞赛关系抽取任务DuIE 2.0基线系统,通过设计结构化标注体系,实现基于ERNIE的端到端SPO抽取模型。 -
事件抽取 DuEE_baseline 语言与智能技术竞赛事件抽取任务DuEE 1.0基线系统,实现基于ERNIE+CRF的Pipeline事件抽取模型。 -
实体链指 DuEL_Baseline 面向中文短文本的实体链指任务(CCKS 2020)的基线系统,实现基于ERNIE和多任务机制的实体链指模型。 -
辅助诊断 SignOrSymptom_Relationship 针对EMR具有无结构化文本和结构化信息并存的特点,结合医疗NLU,以深度学习模型实现EMR的向量化表示、诊断预分类和概率计算。 -
文档级关系抽取 SSAN 引入并建模实体间的依赖结构,在文档级关系抽取任务上取得学界领先效果。 https://arxiv.org/abs/2102.10249

时空数据挖掘

任务类型 目录 简介 论文链接
固定资产价值估计 MONOPOLY 实用的POI商业智能算法,对大量其他的固定资产进行价值估计,包括城市居民对不同公共资产价格评估、私有房价评估偏好的发现与量化分析,以及对评估固定资产价格需考虑的空间范围的确定。 https://dl.acm.org/doi/10.1145/3357384.3357810
兴趣点生成 P3AC 具备个性化的前缀嵌入的POI自动生成。 -
区域生成 P3AC 基于路网进行区域划分的方法, 实现对特定区域基于路网的全划分,区域之间无交叠,无空隙,算法支持对全球的区域划分。 -

许可证书

此向导由PaddlePaddle贡献,受Apache-2.0 license许可认证。

research's People

Contributors

anpark avatar aozhang avatar apphpp avatar benfengxu avatar godfanmiao avatar heda-song avatar hongyuli2018 avatar kahitomi avatar kgresearch avatar legendarydan avatar lielinjiang avatar liuzeming01 avatar liyanyanliyanyan avatar liyingying0113 avatar milkfish1988 avatar miraclebiu avatar parap1uie-s avatar reyonren avatar sserdoubleh avatar steffy-zxf avatar tata1661 avatar weili-nlp avatar wutianyirosun avatar wwqydy avatar xfcygaocan avatar xiao-congxi avatar xiaoguanghu01 avatar xuanmeng-zhang avatar zezeze97 avatar zubingou avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

research's Issues

RuntimeError: parallel_for failed: no kernel image is available for execution on the device

GrahpSum is very intriguing, but I'm unable to test with ./scripts/predict_graphsum_local_multinews.sh.

The log/lanch.log file shows this error: RuntimeError: parallel_for failed: no kernel image is available for execution on the device

+ source ./env_local/env_local.sh
++ set -xe
+++ hostname -i
++ export iplist=127.0.1.1
++ iplist=127.0.1.1
++ unset http_proxy
++ unset https_proxy
+ source ./env_local/utils.sh
++ set -u
+ source ./model_config/graphsum_model_conf_local_multinews
++ task=GraphSum_MDS
++ VOCAB_PATH=./vocab/spm9998_3.model
++ CONFIG_PATH=./model_config/graphsum_config.json
++ TASK_DATA_PATH=/home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle
++ lr_scheduler=noam_decay
++ use_fp16=False
++ use_fuse=True
++ use_hierarchical_allreduce=True
++ nccl_comm_num=3
++ loss_scaling=12800
++ WARMUP_PROP=0.01
++ WARMUP_STEPS=8000
++ beta1=0.9
++ beta2=0.998
++ eps=1e-9
++ LR_RATE=2.0
++ WEIGHT_DECAY=0.01
+ export FLAGS_eager_delete_tensor_gb=1.0
+ FLAGS_eager_delete_tensor_gb=1.0
+ export FLAGS_sync_nccl_allreduce=1
+ FLAGS_sync_nccl_allreduce=1
+ export FLAGS_fraction_of_gpu_memory_to_use=0.98
+ FLAGS_fraction_of_gpu_memory_to_use=0.98
+ export CUDA_VISIBLE_DEVICES=0
+ CUDA_VISIBLE_DEVICES=0
+ python -u ./src/run.py --model_name graphsum --use_cuda true --is_distributed false --use_multi_gpu_test False --use_fast_executor true --use_fp16 False --use_dynamic_loss_scaling False --init_loss_scaling 12800 --weight_sharing true --do_train false --do_val false --do_test true --do_dec true --verbose true --batch_size 30000 --in_tokens true --stream_job '' --init_pretraining_params '' --train_set /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/train --dev_set /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/valid --test_set /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/test --vocab_path ./vocab/spm9998_3.model --config_path model_config/graphsum_config.json --checkpoints ./models/graphsum_multinews --init_checkpoint ./models/graphsum_multinews/step_42976 --decode_path ./results/graphsum_multinews --lr_scheduler noam_decay --save_steps 10000 --weight_decay 0.01 --warmup_steps 8000 --validation_steps 20000 --epoch 100 --max_para_num 30 --max_para_len 60 --max_tgt_len 300 --max_out_len 300 --min_out_len 200 --beam_size 5 --graph_type similarity --len_penalty 0.6 --block_trigram True --report_rouge True --learning_rate 2.0 --skip_steps 100 --grad_norm 2.0 --pos_win 2.0 --label_smooth_eps 0.1 --num_iteration_per_drop_scope 10 --log_file log/graphsum_multinews_test.log --random_seed 1

(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ cat log/lanch.log
-----------  Configuration Arguments -----------
batch_size: 30000
beam_size: 5
beta1: 0.9
beta2: 0.998
block_trigram: True
checkpoints: ./models/graphsum_multinews
config_path: model_config/graphsum_config.json
decode_path: ./results/graphsum_multinews
decr_every_n_nan_or_inf: 2
decr_ratio: 0.8
dev_set: /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/valid
do_dec: True
do_lower_case: True
do_test: True
do_train: False
do_val: False
encoder_json_file: roberta_config/encoder.json
epoch: 100
eps: 1e-09
ernie_config_path: ernie_config/ernie_config.json
ernie_vocab_file: ernie_config/vocab.txt
evaluate_blue: False
grad_norm: 2.0
graph_type: similarity
in_tokens: True
incr_every_n_steps: 100
incr_ratio: 2.0
init_checkpoint: ./models/graphsum_multinews/step_42976
init_loss_scaling: 12800.0
init_pretraining_params:
is_distributed: False
label_smooth_eps: 0.1
learning_rate: 2.0
len_penalty: 0.6
log_file: log/graphsum_multinews_test.log
lr_scheduler: noam_decay
max_out_len: 300
max_para_len: 60
max_para_num: 30
max_seq_len: 512
max_tgt_len: 300
metrics: True
min_out_len: 200
model_name: graphsum
num_iteration_per_drop_scope: 10
pos_win: 2.0
random_seed: 1
report_rouge: True
roberta_config_path: roberta_config/roberta_config.json
roberta_vocab_file: roberta_config/vocab.txt
save_steps: 10000
skip_steps: 100
stream_job:
test_set: /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/test
train_set: /home/matt/mr/algos/GraphSum/data/MultiNews_data_tfidf_30_paddle/train
use_cuda: True
use_dynamic_loss_scaling: False
use_fast_executor: True
use_fp16: False
use_interval: False
use_multi_gpu_test: False
validation_steps: 20000
verbose: True
vocab_bpe_file: roberta_config/vocab.bpe
vocab_path: ./vocab/spm9998_3.model
warmup_proportion: 0.1
warmup_steps: 8000
weight_decay: 0.01
weight_sharing: True
------------------------------------------------
attention_probs_dropout_prob: 0.1
dec_graph_layers: 8
dec_word_pos_embedding_name: dec_word_pos_embedding
enc_graph_layers: 2
enc_sen_pos_embedding_name: enc_sen_pos_embedding
enc_word_layers: 6
enc_word_pos_embedding_name: enc_word_pos_embedding
hidden_act: relu
hidden_dropout_prob: 0.1
hidden_size: 256
initializer_range: 0.02
max_position_embeddings: 512
num_attention_heads: 8
postprocess_command: da
preprocess_command: n
word_embedding_name: word_embedding
------------------------------------------------
[2020-10-23 10:20:10,392 INFO] {'BOS': 4, 'EOS': 5, 'PAD': 6, 'EOT': 3, 'EOP': 7, 'EOQ': 8, 'UNK': 0}
[2020-10-23 10:20:10,393 WARNING] paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[2020-10-23 10:20:11,079 INFO] args.is_distributed: False
W1023 10:20:11.511365 3292334 device_context.cc:236] Please NOTE: device: 0, CUDA Capability: 86, Driver API Version: 11.1, Runtime API Version: 10.0
W1023 10:20:11.512548 3292334 device_context.cc:244] device: 0, cuDNN Version: 8.0.
W1023 10:20:11.862699 3292334 operator.cc:179] truncated_gaussian_random raises an exception thrust::system::system_error, parallel_for failed: no kernel image is available for execution on the device
/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/paddle/fluid/executor.py:779: UserWarning: The following exception is not an EOF exception.
  "The following exception is not an EOF exception.")
Traceback (most recent call last):
  File "./src/run.py", line 35, in <module>
    run_graphsum(args)
  File "/home/matt/mr/algos/Research/NLP/ACL2020-GraphSum/src/networks/graphsum/run_graphsum.py", line 219, in main
    exe.run(startup_prog)
  File "/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/paddle/fluid/executor.py", line 780, in run
    six.reraise(*sys.exc_info())
  File "/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/six.py", line 703, in reraise
    raise value
  File "/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/paddle/fluid/executor.py", line 775, in run
    use_program_cache=use_program_cache)
  File "/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/paddle/fluid/executor.py", line 822, in _run_impl
    use_program_cache=use_program_cache)
  File "/home/matt/anaconda3/envs/graphsum/lib/python3.6/site-packages/paddle/fluid/executor.py", line 899, in _run_program
    fetch_var_name)
RuntimeError: parallel_for failed: no kernel image is available for execution on the device

(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ nvidia-smi
Fri Oct 23 10:21:38 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.23.05    Driver Version: 455.23.05    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 3090    Off  | 00000000:21:00.0 Off |                  N/A |
| 30%   32C    P0    62W / 350W |      0MiB / 24265MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ python -V
Python 3.6.9 :: Anaconda, Inc.
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep nltk
nltk             3.4.5
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep numpy
numpy            1.18.1
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep paddlepaddle
paddlepaddle-gpu 1.6.3.post107
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep pyrouge
pyrouge          0.1.3
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep regex
regex            2020.2.20
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep requests
requests         2.22.0
(graphsum) matt@DeepWhite:~/mr/algos/GraphSum $ pip list | grep sentencepiece
sentencepiece    0.1.85

(graphsum) matt@DeepWhite:~/mr/algos/GraphSum/log $ cat graphsum_multinews_test.log
[2020-10-23 10:16:45,647 INFO] {'BOS': 4, 'EOS': 5, 'PAD': 6, 'EOT': 3, 'EOP': 7, 'EOQ': 8, 'UNK': 0}
[2020-10-23 10:16:45,647 WARNING] paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[2020-10-23 10:16:46,320 INFO] args.is_distributed: False
[2020-10-23 10:19:38,037 INFO] {'BOS': 4, 'EOS': 5, 'PAD': 6, 'EOT': 3, 'EOP': 7, 'EOQ': 8, 'UNK': 0}
[2020-10-23 10:19:38,037 WARNING] paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[2020-10-23 10:19:38,740 INFO] args.is_distributed: False
[2020-10-23 10:20:10,392 INFO] {'BOS': 4, 'EOS': 5, 'PAD': 6, 'EOT': 3, 'EOP': 7, 'EOQ': 8, 'UNK': 0}
[2020-10-23 10:20:10,393 WARNING] paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[2020-10-23 10:20:11,079 INFO] args.is_distributed: False
[2020-10-23 10:43:43,574 INFO] {'BOS': 4, 'EOS': 5, 'PAD': 6, 'EOT': 3, 'EOP': 7, 'EOQ': 8, 'UNK': 0}
[2020-10-23 10:43:43,574 WARNING] paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[2020-10-23 10:43:44,258 INFO] args.is_distributed: False

关于GraphSum数据迁移的问题

@ZeyuChen @kahitomi
你好,很感谢你们如此优秀的开源。
我想问个问题,这份代码迁移到中文数据集的话,sentencepiece vocab file:spm9998_3.model 需要我们自己重新训练吗?直接用你们开源出来的可以吗?

dusql-baseline

运行baseline时, 运行至2020-06-01 15:04:33,010-INFO: value feature is being used这个日志时,再无日志输入,gpu利用率为11m,运行的时lstm编码

PLATO fine-tuning raises error.. Please help!

Hello,
I installed all requirements by using pip install -r requirement.txt
my cuda version, cudnn version are fit for paddlepaddle-gpu == 1.6.1.post107 (which is on your requirements.txt)
However, I met this error when I try to fine-tune the pre-trained model.

Traceback (most recent call last):
File "./run.py", line 23, in
import paddle.fluid as fluid
File "/home/joy/Research/NLP/Dialogue-PLATO/vplato/lib/python3.6/site-packages/paddle/fluid/init.py", line 35, in
from . import framework
File "/home/joy/Research/NLP/Dialogue-PLATO/vplato/lib/python3.6/site-packages/paddle/fluid/framework.py", line 35, in
from . import core
File "/home/joy/Research/NLP/Dialogue-PLATO/vplato/lib/python3.6/site-packages/paddle/fluid/core.py", line 187, in
raise e
File "/home/joy/Research/NLP/Dialogue-PLATO/vplato/lib/python3.6/site-packages/paddle/fluid/core.py", line 167, in
from .core_avx import *
ImportError: /home/joy/Research/NLP/Dialogue-PLATO/vplato/lib/python3.6/site-packages/paddle/fluid/../libs/libmklml_intel.so: symbol __kmpc_omp_task_with_deps, version VERSION not defined in file libiomp5.so with link time reference

  • [[ false = true ]]

I tried my best to fix it, but it didn't work. What is the solution for this problem?

AttributeError: 'Set' Object has no attribute 'append'

I am preparing AICity 2020 data and I have followed the instruction provided on process_aicity_data script. However, I am facing this issue and I've tried to solve it but I couldn't figure it out. Please help me to solve this issue.
PS I am a very new to this. I would really appreciate it.

Traceback (most recent call last):
File "2_prepare_real_trainlist.py", line 37, in
all_ids.append(vid)
AttributeError: 'set' object has no attribute 'append'

image

关于PLATO模型:为什么需要隐变量?

按照文章的介绍,是为了更好地进行一对多生成,但事实上seq2seq模型本身就可以通过采样生成(而不是beam search确定性生成),所以原则上seq2seq模型本身就包含了一对多生成能力,文章所说的常规seq2seq不能很好地做一对多生成的断言似乎不能成立。

那么,隐变量的意义何在呢?此外,我没看到关于隐变量的正则项,那么如何保证隐变量的分布不会退化为一个one hot分布呢(即变成只有一个类,等价于没有隐变量)?

ACL2019_DuConv generative_paddle loss error

Hi,
I want to run the model in generative_paddle, but there is an error when I run run_train.sh:
AssertionError: The loss.shape should be (1L,), but the current loss.shape is (-1L,). Maybe that you should call fluid.layers.mean to process the current loss.
How can I solve it? THANKS!

Research/CV/PaddleReid/process_aicity_data/

2_prepare_real_trainlist.py
37行 all_ids.append(vid) 应该有错误,set不支持append方法,应该是add方法

2_prepare_syn_trainlist.py
29,30 行
color = s.attributes['colorID'].value
cartype = s.attributes['typeID'].value
但是不是所有的车辆都有colorID这个属性,会造成程序崩溃

About ACL2020-GraphSum

Hi, thx for your nice work accepted by acl2020. I check out the link attached on paper leveraging graph to improve abstractive multi-document summarization. But I did not find code and result in this repo and any other branch. How could I access codes and results.

About ACL2019 ARNOR

你好,关于ACL2019收录论文,ARNOR,请问github中给出的的data version2的F1评测结果是macro的计算方式还是micro的计算方式,评测指标是否会去除None的标注。

Can dataset processed in MMPMS be Shared?

MMPMS("Generating Multiple Diverse Responses with Multi-Mapping and Posterior Mapping Selection") evaluate the proposed model on two public conversation dataset: Weibo [Shang et al., 2015] and Reddit [Zhou et al., 2018] that maintain a large repository of post-response pairs from popular social websites.
The paper mentioned that "After basic data cleaning, we have above 2 million pairs in both datasets."
I would like to train the MMPMS from scratch. Could you please share the cleaned data?

InvalidArgumentError

在aistudio上运行,报错误InvalidArgumentError: Python object is not type of St10shared_ptrIN6paddle10imperative7VarBaseEE (at /paddle/paddle/fluid/pybind/imperative.cc:216),该如何解决

DuEL_Baseline 在运行predict.sh的时候,报 var read_file_0.tmp_3 not in this block 错误

详情log如下:
请问是哪里出错了吗,训练是没有问题的

环境:Cuda10+cudnn 7.4 + paddle1.7.1

Traceback (most recent call last):
File "./ernie/infer_type_ranker.py", line 358, in
main(args)
File "./ernie/infer_type_ranker.py", line 163, in main
main_program=predict_prog,
File "/home/hadoop-aipnlp/cephfs/data/gaojianwei/research/ccks2020/DuEL_Baseline/env2/lib/python2.7/site-packages/paddle/fluid/io.py", line 1221, in save_inference_model
prepend_feed_ops(main_program, feeded_var_names)
File "/home/hadoop-aipnlp/cephfs/data/gaojianwei/research/ccks2020/DuEL_Baseline/env2/lib/python2.7/site-packages/paddle/fluid/io.py", line 1031, in prepend_feed_ops
out = global_block.var(name)
File "/home/hadoop-aipnlp/cephfs/data/gaojianwei/research/ccks2020/DuEL_Baseline/env2/lib/python2.7/site-packages/paddle/fluid/framework.py", line 2280, in var
raise ValueError("var %s not in this block" % name)
ValueError: var read_file_0.tmp_3 not in this block

No such file or directory: 'log/graphsum.log'

When I run file 'predict_graphsum_local_multinews.sh' on google colab. I have this problem the following:
Traceback (most recent call last):
File "/content/drive/MyDrive/DATN/GraphSum/src/run.py", line 31, in
init_logger(args.log_file)
File "/content/drive/MyDrive/DATN/GraphSum/src/utils/logging.py", line 33, in init_logger
file_handler = logging.FileHandler(log_file)
File "/usr/lib/python3.6/logging/init.py", line 1032, in init
StreamHandler.init(self, self._open())
File "/usr/lib/python3.6/logging/init.py", line 1061, in _open
return open(self.baseFilename, self.mode, encoding=self.encoding)
FileNotFoundError: [Errno 2] No such file or directory: '/content/log/graphsum.log'

Plato能否在中文语料上从头训练?

您好,我发现Plato还没有chinese版本。我想要在自己的中文数据集上使用plato模型,请问能否从头开始训练?能的话应该如何训练?直接使用英文预训练的checkpoint肯定不行吧。

谢谢

DuIE_Baseline编译报错

执行的时候报错,麻烦看下是啥问题?
Python Call Stacks (More useful to users):

File "/home/yong-group/.local/lib/python3.7/site-packages/paddle/fluid/framework.py", line 2610, in append_op
attrs=kwargs.get("attrs", None))
File "/home/yong-group/.local/lib/python3.7/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
return self.main_program.current_block().append_op(*args, **kwargs)
File "/home/yong-group/.local/lib/python3.7/site-packages/paddle/fluid/layers/sequence_lod.py", line 1057, in sequence_unpad
outputs={'Out': out})
File "/home/yong-group/XYN/PAP挑战赛/DuIE_Baseline/ernie/finetune/relation_extraction_multi_cls.py", line 84, in create_model
lod_labels = fluid.layers.sequence_unpad(labels, seq_lens)
File "/home/yong-group/XYN/PAP挑战赛/DuIE_Baseline/ernie/run_duie.py", line 161, in main
ernie_config=ernie_config)
File "/home/yong-group/XYN/PAP挑战赛/DuIE_Baseline/ernie/run_duie.py", line 411, in
main(args)
InvalidArgumentError: The shape of Input(Length) should be [batch_size]. But received (2)
[Hint: Expected len_dims.size() == 1, but received len_dims.size():2 != 1:1.] at (/paddle/paddle/fluid/operators/sequence_ops/sequence_unpad_op.cc:52)
[operator < sequence_unpad > error]

A question about ACL2020 GraphSum paper

Hi,

Thanks for releasing the code. I read the paper of GraphSum and here is my question.
In section 3.3 You said, "However, thanks to the graph modeling, our model can process much longer inputs." So, how do you process longer inputs? I am very interested in it.

thank you!

运行PLATO模型,训练时进程被Killed

您好,
我在PaddlePaddle 1.6.0,8核8G内存的Linux机器上运行PLATO模型时,报如下错误,请帮忙看看,谢谢!

$ bash scripts/DailyDialog/train.sh

  • SAVE_DIR=outputs/DailyDialog
  • VOCAB_PATH=model/Bert/vocab.txt
  • DATA_DIR=data/DailyDialog
  • INIT_CHECKPOINT=model/PLATO
  • DATA_TYPE=multi
  • USE_VISUALDL=false
  • export CUDA_VISIBLE_DEVICES=
  • CUDA_VISIBLE_DEVICES=
  • export FLAGS_fraction_of_gpu_memory_to_use=0.1
  • FLAGS_fraction_of_gpu_memory_to_use=0.1
  • export FLAGS_eager_delete_scope=True
  • FLAGS_eager_delete_scope=True
  • export FLAGS_eager_delete_tensor_gb=0.0
  • FLAGS_eager_delete_tensor_gb=0.0
  • python -u ./preprocess.py --vocab_path model/Bert/vocab.txt --data_dir data/DailyDialog --data_type multi
  • [[ false = true ]]
  • python -u ./run.py --do_train true --vocab_path model/Bert/vocab.txt --data_dir data/DailyDialog --data_type multi --batch_size 6 --valid_steps 2000 --num_type_embeddings 2 --use_discriminator true --num_epoch 20 --lr 1e-5 --save_checkpoint false --save_summary false --init_checkpoint model/PLATO --save_dir outputs/DailyDialog
    {
    "do_train": true,
    "do_test": false,
    "do_infer": false,
    "num_infer_batches": null,
    "hparams_file": null,
    "BPETextField": {
    "vocab_path": "model/Bert/vocab.txt",
    "filtered": false,
    "max_len": 256,
    "min_utt_len": 1,
    "max_utt_len": 50,
    "min_ctx_turn": 1,
    "max_ctx_turn": 16,
    "max_knowledge_num": 16,
    "max_knowledge_len": 16,
    "tokenizer_type": "Bert"
    },
    "Dataset": {
    "data_dir": "data/DailyDialog",
    "data_type": "multi"
    },
    "Trainer": {
    "use_data_distributed": false,
    "valid_metric_name": "-loss",
    "num_epochs": 20,
    "save_dir": "outputs/DailyDialog",
    "batch_size": 6,
    "log_steps": 100,
    "valid_steps": 2000,
    "save_checkpoint": false,
    "save_summary": false,
    "shuffle": true,
    "sort_pool_size": 0
    },
    "Model": {
    "init_checkpoint": "model/PLATO",
    "model": "UnifiedTransformer",
    "num_token_embeddings": -1,
    "num_pos_embeddings": 512,
    "num_type_embeddings": 2,
    "num_turn_embeddings": 16,
    "num_latent": 20,
    "tau": 0.67,
    "with_bow": true,
    "hidden_dim": 768,
    "num_heads": 12,
    "num_layers": 12,
    "padding_idx": 0,
    "dropout": 0.1,
    "embed_dropout": 0.0,
    "attn_dropout": 0.1,
    "ff_dropout": 0.1,
    "use_discriminator": true,
    "dis_ratio": 1.0,
    "weight_sharing": true,
    "pos_trainable": true,
    "two_layer_predictor": false,
    "bidirectional_context": true,
    "label_smooth": 0.0,
    "initializer_range": 0.02,
    "lr": 1e-05,
    "weight_decay": 0.0,
    "max_grad_norm": null
    },
    "Generator": {
    "generator": "BeamSearch",
    "min_gen_len": 1,
    "max_gen_len": 30,
    "beam_size": 5,
    "length_average": false,
    "length_penalty": -1.0,
    "ignore_unk": true
    }
    }
    Loading parameters from model/PLATO
    Loaded parameters from model/PLATO
    scripts/DailyDialog/train.sh: line 45: 2041 Killed python -u ./run.py --do_train true --vocab_path $VOCAB_PATH --data_dir $DATA_DIR --data_type $DATA_TYPE --batch_size 6 --valid_steps 2000 --num_type_embeddings 2 --use_discriminator true --num_epoch 20 --lr 1e-5 --save_checkpoint false --save_summary $USE_VISUALDL --init_checkpoint $INIT_CHECKPOINT --save_dir $SAVE_DIR
  • [[ false = true ]]

errors running DuReader-Robust-BASELINE

W0414 16:45:31.025034 10900 device_context.cc:237] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.0, Runtime API Version: 9.0
W0414 16:45:31.028314 10900 device_context.cc:245] device: 0, cuDNN Version: 7.6.

I0414 16:45:33.431712 10900 parallel_executor.cc:440] The Program will be executed on CUDA using ParallelExecutor, 2 cards are used, so 2 programs are executed in parallel.
W0414 16:45:36.668439 10900 init.cc:209] Warning: PaddlePaddle catches a failure signal, it may not work properly
W0414 16:45:36.668462 10900 init.cc:211] You could check whether you killed PaddlePaddle thread/process accidentally or report the case to PaddlePaddle
W0414 16:45:36.668467 10900 init.cc:214] The detail failure signal is:

W0414 16:45:36.668470 10900 init.cc:217] *** Aborted at 1586853936 (unix time) try "date -d @1586853936" if you are using GNU date ***
W0414 16:45:36.670137 10900 init.cc:217] PC: @ 0x0 (unknown)
W0414 16:45:36.670218 10900 init.cc:217] *** SIGSEGV (@0x0) received by PID 10900 (TID 0x7f7fa99b1700) from PID 0; stack trace: ***
W0414 16:45:36.671615 10900 init.cc:217] @ 0x7f7fa959d390 (unknown)
W0414 16:45:36.673030 10900 init.cc:217] @ 0x0 (unknown)

关于EMNLP2019-AKGCM中,Minerva应用在对话中的query

您好,

在Minerva文章中解决的是三元组的QA问题,所以他们的query也来自于relation。
那对于对话中的query,是第一个人说的话吗?
那训练的数据集是将triple中的relation全部换成sentence吗?

谢谢!

InvalidArgumentError: The shape of input[0] and input[1] is expected to be equal.But received input[0]'s shape = [-1, 0, 1], input[1]'s shape = [-1, 1, 1, 1].

I'm trying to test the GraphSum model with the command ./scripts/predict_graphsum_local_multinews.sh found in the documentation, but I get this error:

Error Message Summary:
----------------------
InvalidArgumentError: The shape of input[0] and input[1] is expected to be equal.But received input[0]'s shape = [-1, 0, 1], input[1]'s shape = [-1, 1, 1, 1].
  [Hint: Expected inputs_dims[i].size() == out_dims.size(), but received inputs_dims[i].size():4 != out_dims.size():3.] at (/paddle/paddle/fluid/operators/concat_op.h:40)
  [operator < concat > error]

I can't really debug this issue in PyCharm because of all the shell scripts involved. Any advice would be greatly appreciated. Thanks.

ACL2019-ARNOR dataset clarification

I have the following questions on data version 2.0.0.

  1. Is the dev.json file used for validation? or this is just another test set? did you use a part of train.json for validation ?
  2. did you include the instances in dev.json and test.json which is marked as is_noise=true in the F1 score calculation ?
  3. How many relations are used for the experiments ? Can you please provide a list of them ?

Thanks !!!!

NLP/ACL2018_DuReader ModuleNotFoundError: No module named 'bidaf_model'

image
尝试运行PaddlePaddle/Research/NLP/ACL2018_DuReader, 执行到'评估'阶段时,run.py中导包失败

Traceback (most recent call last):
  File "run.py", line 41, in <module>
    import bidaf_model as rc_model
ModuleNotFoundError: No module named 'bidaf_model'

检查后,没有在项目中找到bidaf_model相关代码。
希望可以帮忙解决一下,谢谢。

2_prepare_syn_trainlist

Traceback (most recent call last):
File "2_prepare_syn_trainlist.py", line 29, in
color = s.attributes['colorID'].value
File "/home/eini/anaconda3/lib/python3.7/xml/dom/minidom.py", line 552, in getitem
return self._attrs[attname_or_tuple]
KeyError: 'colorID'

aistudio上 使用predict.sh 预测,报错 ValueError: var read_file_0.tmp_3 not in this block

环境:aistudio Cuda9.2+cudnn 7.6 + paddle1.8.4
使用paddle1.5就没事,但是这个就不行,请问怎么解决

报错log如下

2020-11-23 14:35:49,888-WARNING: paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
[WARNING] 2020-11-23 14:35:49,888 [       io.py:  712]:	paddle.fluid.layers.py_reader() may be deprecated in the near future. Please use paddle.fluid.io.DataLoader.from_generator() instead.
----------place-----------
CUDAPlace(0)
W1123 14:35:50.949314   187 device_context.cc:252] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 9.0
W1123 14:35:51.145216   187 device_context.cc:260] device: 0, cuDNN Version: 7.6.
2020-11-23 14:35:57,108-INFO: Load pretraining parameters from ./checkpoints/step_20000.
[INFO] 2020-11-23 14:35:57,108 [     init.py:  101]:	Load pretraining parameters from ./checkpoints/step_20000.
2020-11-23 14:35:57,108-INFO: save inference model to ./checkpoints/inference_model/step_20000_inference_model
[INFO] 2020-11-23 14:35:57,108 [infer_type_ranker.py:  158]:	save inference model to ./checkpoints/inference_model/step_20000_inference_model
Traceback (most recent call last):
  File "./ernie/infer_type_ranker.py", line 358, in <module>
    main(args)
  File "./ernie/infer_type_ranker.py", line 163, in main
    main_program=predict_prog
  File "/opt/conda/envs/python27-paddle120-env/lib/python2.7/site-packages/paddle/fluid/io.py", line 1247, in save_inference_model
    prepend_feed_ops(main_program, feeded_var_names)
  File "/opt/conda/envs/python27-paddle120-env/lib/python2.7/site-packages/paddle/fluid/io.py", line 1043, in prepend_feed_ops
    out = global_block.var(name)
  File "/opt/conda/envs/python27-paddle120-env/lib/python2.7/site-packages/paddle/fluid/framework.py", line 2377, in var
    raise ValueError("var %s not in this block" % name)
ValueError: var read_file_0.tmp_3 not in this block

Why the test result of Multi-News only has 5590 lines?

Hi,

Thanks for releasing the result of test set.
I'm doubting why the test result only has 5590 lines? The original Multi-News dataset contains 5622 document-pairs for testing. Did you exclude the outliers? Did you do the same during training?
Hope get your reply soon.

Joyce

Duplicates in the ACL2020_SignOrSymptom_Relationship

Thanks for publishing the KG!

It seems that there are duplicates in the disease-finding relations. For example,

支气管哮喘 气喘 Symptom

appeared at least twice in the relations_respiration_all.txt. Would this matter to the results?

Best wishes,
A

多卡运行NLP/DuReader-Robust-BASELINE报错:Tensor holds no memory. Call Tensor::mutable_data first.

NLP/DuReader-Robust-BASELINE的训练程序,单卡时正常运行,多卡时则会报错,具体信息如下:

C++ Call Stacks (More useful to developers):

0 std::string paddle::platform::GetTraceBackString<std::string const&>(std::string const&, char const*, int)
1 paddle::platform::EnforceNotMet::EnforceNotMet(std::string const&, char const*, int)
2 paddle::framework::Tensor::check_memory_size() const
3 long const* paddle::framework::Tensor::data() const
4 paddle::operators::LookupTableV2CUDAKernel::Compute(paddle::framework::ExecutionContext const&) const
5 std::Function_handler<void (paddle::framework::ExecutionContext const&), paddle::framework::OpKernelRegistrarFunctor<paddle::platform::CUDAPlace, false, 0ul, paddle::operators::LookupTableV2CUDAKernel, paddle::operators::LookupTableV2CUDAKernel, paddle::operators::LookupTableV2CUDAKernelpaddle::platform::float16 >::operator()(char const*, char const*, int) const::{lambda(paddle::framework::ExecutionContext const&)#1}>::M_invoke(std::Any_data const&, paddle::framework::ExecutionContext const&)
6 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void
, boost::detail::variant::void
, boost::detail::variant::void
, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&, paddle::framework::RuntimeContext*) const
7 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&) const
8 paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&)
9 paddle::framework::details::ComputationOpHandle::RunImpl()
10 paddle::framework::details::ThreadedSSAGraphExecutor::RunOpSync(paddle::framework::details::OpHandleBase*)
11 std::_Function_handler<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> (), std::__future_base::_Task_setter<std::unique_ptr<std::__future_base::_Result, std::__future_base::_Result_base::_Deleter>, void> >::_M_invoke(std::_Any_data const&)
12 std::__future_base::_State_base::_M_do_set(std::function<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> ()>&, bool&)
13 ThreadPool::ThreadPool(unsigned long)::{lambda()#1}::operator()() const


Python Call Stacks (More useful to users):

File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/framework.py", line 2459, in append_op
attrs=kwargs.get("attrs", None))
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
return self.main_program.current_block().append_op(*args, **kwargs)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/input.py", line 268, in embedding
'padding_idx': padding_idx
File "/home/zhan/Research-master/NLP/DuReader-Robust-BASELINE/src/model/ernie.py", line 97, in _build_model
name=self._pos_emb_name, initializer=self._param_initializer))
File "/home/zhan/Research-master/NLP/DuReader-Robust-BASELINE/src/model/ernie.py", line 81, in init
self.build_model(src_ids, position_ids, sentence_ids, input_mask)
File "", line 39, in create_model
use_fp16=args.use_fp16)
File "", line 6, in
is_training=True)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3331, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3254, in run_ast_nodes
if (await self.run_code(code, result, async
=asy)):
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3063, in run_cell_async
interactivity=interactivity, compiler=compiler, result=result)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/async_helpers.py", line 68, in _pseudo_sync_runner
coro.send(None)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 2886, in _run_cell
return runner(coro)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 2858, in run_cell
raw_cell, store_history, silent, shell_futures)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/zmqshell.py", line 536, in run_cell
return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/ipkernel.py", line 300, in do_execute
res = shell.run_cell(code, store_history=store_history, silent=silent)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 209, in wrapper
yielded = next(result)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/kernelbase.py", line 545, in execute_request
user_expressions, allow_stdin,
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 209, in wrapper
yielded = next(result)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/kernelbase.py", line 268, in dispatch_shell
yield gen.maybe_future(handler(stream, idents, msg))
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 209, in wrapper
yielded = next(result)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/kernelbase.py", line 365, in process_one
yield gen.maybe_future(dispatch(*args))
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 748, in run
yielded = self.gen.send(value)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 714, in init
self.run()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 225, in wrapper
runner = Runner(result, future, yielded)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/kernelbase.py", line 381, in dispatch_queue
yield self.process_one()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 748, in run
yielded = self.gen.send(value)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/gen.py", line 787, in inner
self.run()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/ioloop.py", line 743, in _run_callback
ret = callback()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/ioloop.py", line 690, in
lambda f: self._run_callback(functools.partial(callback, future))
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/asyncio/events.py", line 88, in _run
self._context.run(self._callback, *self._args)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/asyncio/base_events.py", line 1786, in _run_once
handle._run()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/asyncio/base_events.py", line 541, in run_forever
self._run_once()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/tornado/platform/asyncio.py", line 149, in start
self.asyncio_loop.run_forever()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel/kernelapp.py", line 583, in start
self.io_loop.start()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/traitlets/config/application.py", line 664, in launch_instance
app.start()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/site-packages/ipykernel_launcher.py", line 16, in
app.launch_new_instance()
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home/zhan/anaconda3/envs/paddle/lib/python3.7/runpy.py", line 193, in _run_module_as_main
"main", mod_spec)


Error Message Summary:

PaddleCheckError: holder_ should not be null
Tensor holds no memory. Call Tensor::mutable_data first. at [/paddle/paddle/fluid/framework/tensor.cc:23]
[operator < lookup_table_v2 > error]

运行环境为:
paddlepaddle-gpu 1.6.1.post107
cuda 10.0
cudnn 7.6.4
nccl 2.6.4

请问有预训练用的数据吗?

就是原文中提到的 Large-scale conversation datasets – Twitter (Cho et al., 2014) and Reddit (Zhou et al., 2018; Galley et al., 2019) are employed for pretraining, which results in 8.3 million training samples in total.

About ACL2020-GraphSum

您好!我想用自己的数据测试模型,请问像WIKI.test.0.json这样的数据是怎么生成的呢?有相应程序吗?

I think parallel.py in PLATO implementation raises attribute error

Hello,
It might be a silly question because it's my first time using paddle based codes... I hope your understanding!

I'm trying to run PLATO fine-tuning code, and I met 'Parameter' object has no attribute '_grad_ivar' error' at apply_collectice_grads function in parallel.py in plato implementation.

I also noticed that this function is also implemented in paddle/fluid/dygraph/parallel.py, and it was slightly different from the implementation in plato.

Therefore, I changed the function in plato just like the function in paddle implementaion.
As a result, I can run this code, but I still don't know if this method will make sense... I need your help!!

Thank you.


        for param in self._layers.parameters():
            # NOTE(zcd): The grad_ivar maybe no generated.
            #if param.trainable and param._grad_ivar():
            if param.trainable and param._ivar._grad_ivar():
                g_var = param._grad_ivar()
                grad_vars.append(g_var)
                assert g_var not in grad_var_set
                grad_var_set.add(g_var)

at apply_collectice_grads function in Research/NLP/Dialogue-PLATO/plato/modules/parallel.py


        for param in self._layers.parameters():
            # NOTE(zcd): The grad_ivar maybe no generated.
            if param.trainable and param._ivar._grad_ivar():
                g_var = framework.Variable(
                    block=self._helper.main_program.current_block(),
                    name=param._ivar._grad_name(),
                    stop_gradient=True,
                    ivar=param._ivar._grad_ivar())
                grad_vars.append(g_var)
                assert g_var not in grad_var_set
                grad_var_set.add(g_var)

at apply_collectice_grads function in paddle/fluid/dygraph/parallel.py

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.