Developer Guide
Algorithms
Yolov5
-
Create industrial model
Model extra information
{ "labels": [ "license-plate", "vehicle" ] }
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment data /opt/ml/input/data/cfg/data.yaml Data configuration file cfg yolov5s.yaml Model configuration file weight yolov5s Model weight file (if it is not existed yet and it is standard weight file, it will download automatically project /opt/ml/model/ Model project directory name tutorial Model name img 640 Image size batch 16 Batch size epochs 10 Number of epochs Input data configuration
Channel name Mandatory Comment images Yes S3 URI of images which contains train, validation images labels Yes S3 URI of labels which contains train, validation labels cfg Yes S3 URI of cfg which contains data.yaml weights No S3 URI of weights which contains model file Sample data configuration file
train: /opt/ml/input/data/images/train/ val: /opt/ml/input/data/images/valid/ # number of classes nc: 2 # class names names: ['license-plate','vehicle']
-
Quickstart - deploy
Environment variables
Environment variable Default value Comment model_name custom Indicate if it is a custom model or pre-trained model. In case it is a custom model, the best.pt will be loaded from the specific location. Otherwise, pre-trained model file will be loaded. size 415 Chunk image size to be passed to model -
Quickstart - inference
HTTP request
Inference approach HTTP Body Comment Raw image data Bytes of image file ContentType: image/png, image/jpg, image/jpeg S3 image data { “bucket”: [s3-bucket], “image_uri”: [s3-key] } ContentType: application/json HTTP response
- One row per object
- Each row is class x_center y_center width height format.
- Box coordinates must be normalized by the dimensions of the image (i.e. have values between 0 and 1)
- Class numbers are zero-indexed (start from 0). For example:
1 0.511271 0.571540 0.936967 0.631636 0 0.887925 0.714687 0.072043 0.101056
GluonCV
-
Create industrial model
Model extra information
Image search { "task": "search" }
Image classification { "task": "classification", "classes": [ "tench", "goldfish", ... ] }
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment classes 10 Number of classes batch_size 8 Batch size epochs 20 Number of epochs learning_rate 0.001 Learning rate momentum 0.9 Momentum wd 0.0001 wd num_workers 8 Number of workers model_name ResNet50_v2 Pretrained model name Input data configuration
Channel name Mandatory Comment train Yes S3 URI of train images val Yes S3 URI of val images test Yes S3 URI of test images -
Quickstart - deploy
Image search
Image classification
Environment variables
Environment variable Default value Comment task search Indicate if the task is search or classification classes 1000 Number of classes model_name ResNet50_v2 Pretrained model name -
Quickstart - inference
HTTP request
Inference approach HTTP Body Comment Raw image data Bytes of image file ContentType: image/png, image/jpg, image/jpeg S3 image data { “bucket”: [s3-bucket], “image_uri”: [s3-key] } ContentType: application/json HTTP response
- If task is search, return 2048 dimension embedding vector.
- If task is classification, return top-k matched class id array.
PaddleOCR
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment classes 10 Number of classes batch_size 8 Batch size epochs 20 Number of epochs learning_rate 0.001 Learning rate momentum 0.9 Momentum wd 0.0001 wd num_workers 8 Number of workers model_name ResNet50_v2 Pretrained model name Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train dataset pretrained_models Yes S3 URI of pretrained models -
Quickstart - deploy
Environment variables
Environment variable Default value Comment task ocr Indicate if it is an OCR task or structure task device cpu CPU or GPU device det_model_dir None det model directory when it is an OCR task rec_model_dir None Rec model directory when it is an OCR task table_model_dir None table model directory when it is an OCR task rec_char_dict_path None rec custom dictionary path when it is an structure task table_char_dict_path None table custom dictionary path when it is an structure task lang ch language table True Indicate if table recognition is enabled when it is a structure task layout True Indicate if layout recognition is enabled when it is a structure task ocr True Indicate if ocr recognition is enabled when it is a structure task -
Quickstart - inference
HTTP request
Inference approach HTTP Body Comment Raw image data Bytes of image file ContentType: image/png, image/jpg, image/jpeg S3 image data { “bucket”: [s3-bucket], “image_uri”: [s3-key] } ContentType: application/json HTTP response
- One row per text box
- Each row is in x_left_top, y_left_top, x_right_top, y_right_top, x_right_bottom, y_right_bottom, x_left_bottom, y_left_bottom format, for example:
479,138,1871,138,1871,198,479,198,英开曼群岛商史泰博股份有限公司台灣分公司 972,212,1386,219,1385,283,971,275,電子發票證明聨 1021,304,1336,304,1336,353,1021,353,2022-01-06 57,371,496,371,496,417,57,417,發票号碼:WP79071184 1625,364,1771,364,1771,413,1625,413,格式:25 61,424,646,424,646,470,61,470,實方名:網资讯股份有限公司 54,477,429,477,429,523,54,523,統一编号:24549210 57,530,157,530,157,576,57,576,地址: 54,576,150,576,150,629,54,629,備: 2029,629,2293,629,2293,668,2029,668,第1真/共1真 921,678,1432,678,1432,724,921,724,視同正本,凡經改即無效 1318,731,1400,731,1400,774,1318,774,軍價 1579,731,1671,731,1671,774,1579,774,金额 486,735,568,735,568,777,486,777,品名 1064,735,1154,735,1154,774,1064,774,數量 1986,728,2075,728,2075,781,1986,781,備 1379,781,1493,781,1493,834,1379,834,11.43 57,788,864,788,864,834,57,834,雄狮SIMBALION细字奇翼筆600/黑/1.0mm 1196,788,1225,788,1225,837,1196,837,5 1650,781,1761,781,1761,834,1650,834,57.14 1196,834,1225,834,1225,897,1196,897,3 1357,837,1489,837,1489,887,1357,887,100.00 61,841,807,841,807,887,61,887,立强REGINA無耳三孔D型±/R8603D/黑 1629,834,1761,834,1761,887,1629,887,300.00 1196,887,1225,887,1225,950,1196,950,5 1379,887,1489,887,1489,940,1379,940,56.19 57,894,893,894,893,940,57,940,11孔透明萬能袋/A4/亮面/0.04mm/100张/包 1620,888,1758,879,1762,932,1624,941,280.95 1377,941,1489,931,1494,988,1382,997,15.24 57,947,836,947,836,993,57,993,SDI小三角迥针0731B/25.4mm/70支/盒 1196,947,1225,947,1225,989,1196,989,2 1650,940,1761,940,1761,993,1650,993,30.48 1379,996,1493,996,1493,1046,1379,1046,61.90 54,1000,989,1000,989,1046,54,1046,3M小管芯隐形膠带810/19mmx32.9M/纸盒-3/4时 1196,1000,1225,1000,1225,1042,1196,1042,1 1650,996,1761,996,1761,1046,1650,1046,61.90 57,1053,118,1053,118,1099,57,1099,/卷 1379,1102,1489,1102,1489,1152,1379,1152,57.14 54,1106,986,1106,986,1152,54,1152,3M大管芯OPP透明文具膠带502/18mmx36M/8卷 1196,1109,1225,1109,1225,1144,1196,1144,1 1650,1102,1761,1102,1761,1152,1650,1152,57.14 61,1159,118,1159,118,1201,61,1201,/束 1379,1201,1489,1201,1489,1254,1379,1254,85.72 53,1208,996,1201,997,1250,54,1258,3M大管芯OPP透明封箱膠带/3036-6PK/48mmx36 1629,1201,1761,1201,1761,1250,1629,1250,171.43 1200,1215,1221,1215,1221,1247,1200,1247,2 64,1265,236,1265,236,1303,64,1303,M/6卷/束 1379,1303,1493,1303,1493,1356,1379,1356,10.48 61,1311,857,1303,857,1353,61,1360,得力Deli彩色迥纹針/E39716/33mm/100支 1650,1303,1761,1303,1761,1353,1650,1353,10.48 51,2769,265,2777,263,2830,49,2822,銷售额合计 1678,2776,2278,2773,2279,2822,1679,2826,970營業人用统一發票専用章 125,2829,264,2829,264,2879,125,2879,營業税 450,2833,536,2833,536,2875,450,2875,鹰税 832,2833,957,2833,957,2872,832,2872,零税率 1218,2833,1307,2833,1307,2875,1218,2875,免税 1707,2829,2261,2829,2261,2875,1707,2875,48英曼群岛商史泰博股份 50,2879,136,2879,136,2932,50,2932,總計 1654,2879,2132,2879,2132,2925,1654,2925,1,018有限公司台灣分公司 361,2928,954,2936,953,3010,360,3002,壹仟零壹拾捌元整 1753,2928,2128,2925,2129,2974,1754,2978,統一编号:27946944 54,2939,393,2939,393,2999,54,2999,總計新台 1764,2985,1971,2985,1971,3024,1764,3024,负贵人:吴 1754,3027,2279,3031,2278,3080,1753,3077,地址:新北市新莊區思源路60 1752,3076,1912,3084,1909,3134,1749,3125,1号15楼
CPT
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment model_name_or_path fnlp/cpt-large Pretrained model name num_train_epochs 10 Number of epochs per_device_train_batch_size 4 Batch size per device text_column text Label of text column summary_column summary Label of summary column output_dir /opt/ml/model Output directory train_file /opt/ml/input/data/dataset/train.json Path of train file validation_file /opt/ml/input/data/dataset/val.json Path of validation file test_file /opt/ml/input/data/dataset/test.json Path of test file val_max_target_length 80 Max target length path json Extension name Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train dataset Content format of train/validation/test file
{ “text”: <text>, “summary”: <summary> }
-
Quickstart - deploy
Environment variables
Environment variable Default value Comment input_max_length 512 Maximum effective input length output_max_length 512 Maximum effective output length top_p 0.95 Top probability of output summary -
Quickstart - inference
HTTP request
{ “inputs”: “中国农业银行股份有限公司安远县支行与魏坤元、魏松兰等借款合同纠纷一审民事判决书 江西省安远县人民法院 民 事 判 决 书 (2017)赣0726民初928号原告中国农业银行股份有限公司安远县支行。 住所地:安远县欣山镇龙泉路12号。 法定代表人唐文中,系该行行长。 委托代理人严海,系该行工作人员。 代理权限:代为承认、放弃或者变更诉讼请求,进行和解、提起反诉或者上诉。 被告魏坤元,男,1957年9月9日生,汉族,江西省安远县人,住安远县。 被告魏松兰,男,1971年12月16日生,汉族,江西省安远县人,住安远县。 被告魏碧星,男,1982年1月20日生,汉族,江西省安远县人,住安远县。 原告中国农业银行股份有限公司安远县支行(以下简称农行安远支行)诉被告魏坤元、魏松兰、魏碧星借款合同纠纷一案,本院立案受理后,依法由审判员徐海峰适用简易程序,于2017年8月8日公开开庭进行了审理。 原告农行安远支行的委托代理人严海到庭参加了诉讼,被告魏坤元、魏松兰、魏碧星经本院传票传唤无正当理由未到庭参加诉讼。 本案现已审理终结。 原告农行安远支行诉称,被告魏坤元于2013年12月19日向原告申请农户小额最高额可循环贷款一笔,金额50000元,并由被告魏松兰、魏碧星提供保证担保,合同到期日为2016年12月18日,合同约定在最高额度和期限内,借款人随借随还,自助放款还款,单笔借款期限最长不超过一年。 合同期限内,借款人魏坤元于2015年12月23日通过原告自助电子渠道申请贷款一笔、金额50000元,到期日为2016年11月22日,至今仍结欠原告贷款本金50000元及827.95元(利息计算至2016年12月20日止)。 该笔贷款已逾期,为此,原告诉至法院,请求依法判令:1、被告魏坤元、魏松兰、魏碧星归还原告贷款本金50000元及利息827.95元(利息计算至2016年12月20日止),之后的利息按合同约定的罚息利率计算; 2、本案诉讼费用由被告承担。 被告魏松兰、魏碧星未作答辩。 被告魏坤元未到庭答辩,其在本院的《询问笔录》中辩称,对原告起诉没有异议,确实是和被告魏松兰、魏碧星组成联保小组,相互担保,被告魏碧星是被告魏坤元的儿子,被告魏松兰是被告魏坤元的堂弟。 三被告各向原告借款5万元,拖欠了本息至今。 经审理查明,2013年11月25日,被告魏坤元、魏松兰、魏碧星三人组成联保小组,相互承担连带保证责任向原告农行安远支行申请借款。 三被告组成联保小组后,被告魏坤元向原告借款50000元,并签订了《农户贷款借款合同》(以下简称《借款合同》),合同内容:“第一条借款金额/可循环借款额度(人民币大写):伍万元。 用款方式:自助可循环方式。 自2013年12月19日起至2016年12月18日(额度有效期),借款人可在伍万元的可循环借款额度内向贷款人申请借款,单笔借款期限最长不超过壹年且到期日最迟不得超过额度有效期。 借款用途:生产经营。 第二条本合同项下,借款执行利率以借款发放当日中国人民银行同期同档次人民币贷款基准利率基础上浮30%确定。 1年期以内(含)的借款执行浮动利率。 1年期以上的借款执行浮动利率。 浮动利率指如遇中国人民银行人民币贷款基准利率调整,自基准利率调整之日起,按调整后相应期限档次的基准利率和本合同约定的借款利率浮动幅度确定新的借款执行利率,且不再另行通知借款人和担保人。 第五条保证方式为连带责任保证,保证期间为借款期限届满之日起二年。 第六条借款人未按约定期限归还借款本金,贷款人对逾期借款从逾期之日起在借款执行利率基础上上浮百分之伍拾计收罚息,直至本息清偿为止。 ……” ,三被告均在该《借款合同》上签名捺印。 合同期限内,被告魏坤元于2015年12月23日借到原告农行安远支行发放的借款本金50000元,借款凭证上载明借款金额为伍万元整,执行利率为5.655%,逾期利率为8.4825%,借款日期为2015年12月23日,到期日期为2016年11月22日。 被告魏坤元借款后,经原告多次催款,截至2017年6月20日,仍拖欠原告借款本金50000元及利息2993.36元。 上述事实,有原告的陈述,原告提交的《借款合同》、《中国农业银行借款凭证》、《联合保证担保承诺书》、《中国农业银行农户小额贷款业务申请表》、《中国农业银行农户小额贷款面谈记录》、《本息清单》等证据予以证实,上述证据经庭审审查,能相互印证,本院予以确认。 本院认为,被告魏坤元向原告农行安远支行借款,双方签订了借款合同,原告也依约向被告魏坤元发放了借款,由此形成的借款合同关系合法有效,受法律保护。 借款到期后,被告魏坤元应清偿借款本金并依约支付利息,但其至今未还清借款本金及相应利息,其行为显属违约。 由于被告魏坤元、魏松兰、魏碧星三人组成联保小组,三被告中任一人借款均由其他二被告提供连带责任保证,即三被告相互承担连带保证责任。 因此,被告魏松兰、魏碧星二人应与债务人被告魏坤元承担连带偿还责任,依约偿还原告借款本金及相应利息。 综上,依照《中华人民共和国合同法》第一百零七条、第二百零四条、第二百零五条、第二百零七条,《中华人民共和国担保法》第十二条、第十八条、第二十一条,《中华人民共和国民事诉讼法》第一百四十四条之规定,判决如下:一、被告魏坤元于本判决生效后三十日内偿还原告中国农业银行股份有限公司安远县支行借款本金50000元及利息(截至2017年6月20日的利息为2993.36元,2017年6月20日之后的利息按照合同约定的罚息利率计算)。 二、被告魏松兰、魏碧星对上述款项承担连带偿还责任。 如果未按本判决指定的期间履行给付金钱义务,应当依照《中华人民共和国民事诉讼法》第二百五十三条之规定,加倍支付迟延履行期间的债务利息。 案件受理费1070元,减半收取535元,由被告魏坤元、魏松兰、魏碧星共同负担。 如不服本判决,可在判决书送达之日起十五日内,向本院递交上诉状,并按对方当事人的人数提出副本(在递交上诉状之日起七日内预交上诉费,缴交上诉费账号:99×××88,开户行:招商银行赣州长征大道支行,户名:江西省赣州市中级人民法院,备注栏注明上诉费),上诉于江西省赣州市中级人民法院。 (法律文书生效后,一方拒绝履行的,对方当事人向本院申请执行的期限是从判决书规定的履行期限届满二年内)审判员 徐海峰 二〇一七年九月十一日 书记员 李魁鹏” }
HTTP response
{ “result”: “[SEP] [CLS] 原 被 告 系 借 款 合 同 纠 纷 。 原 告 提 出 诉 讼 请 求 : 被 告 偿 还 原 告 贷 款 及 利 息 、 罚 息 ; 保 证 人 承 担 连 带 清 偿 责 任 。 被 告 未 答 辩 。 经 审 查 , 原 告 与 被 告 签 订 的 农 户 贷 款 借 款 及 担 保 合 同 合 法 有 效 , 被 告 应 当 按 照 合 同 约 定 履 行 偿 还 借 款 义 务 , 否 则 原 告 对 被 告 抵 押 物 享 有 优 先 受 偿 权 。 综 上 , 依 照 《 中 华 人 民 共 和 国 合 同 法 》 第 六 十 条 第 一 款 、 第 一 百 九 十 六 条 、 第 二 百 零 五 条 、 二 百 一 十 一 条 、 《 担 保 法 》 及 《 最 高 人 民 法 院 关 于 适 用 若 干 问 题 的 解 释 ( 二 ) 》 第 二 十 四 条 及 《 民 事 诉 讼 法 》 之 规 定 , 判 决 被 告 给 付 原 告 农 户 最 高 额 可 循 环 贷 款 逾 期 利 息 及 罚 息 。 [SEP]” }
GABSA
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment task uabsa The name of the task, selected from: [uabsa, aste, tasd, aope] dataset rest14 The name of the dataset, selected from: [laptop14, rest14, rest15, rest16] model_name_or_path t5-base Pretrained model name paradigm annotation The way to construct target sentence, selected from: [annotation, extraction] do_train True Whether to run training. do_eval False Whether to run eval on the dev/test set. do_batch_predict False Whether to run batch prediction. do_direct_eval False Whether to run direct eval on the dev/test set. do_direct_predict False Whether to run direct eval on the dev/test set. max_seq_length 128 Maximum sequence length train_batch_size 16 Batch train size eval_batch_size 16 Batch eval size gradient_accumulation_steps 1 Number of updates steps to accumulate before performing a backward/update pass. learning_rate 3e-4 Learning rate num_train_epochs 20 Number of train epochs seed 42 seed ckpoint_path /opt/ml/model/cktepoch=1.ckpt Checkpoint path weight_decay 0.0 Weight decay adam_epsilon 1e-8 Adam epsilon warmup_steps 0.0 Warmup steps Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train dataset Content format of train/validation/test file
- Each row contains origin string####[(word, aspect, sentiment),…], for example: The wine list is interesting and has many good values .####[('wine list', 'drinks style_options', 'positive'), ('wine list', 'drinks prices', 'positive')]
-
Quickstart - deploy
Environment variables
Environment variable Default value Comment input_max_length 512 Maximum effective input length output_max_length 512 Maximum effective output length top_p 0.95 Top probability of output summary -
Quickstart - inference
HTTP request
{ "inputs": "The wine list is wonderful and the food reminds me of my recent trip to Italy ." }
HTTP response
{ "result": "(wine list, drinks style_options, positive); (food, food quality, positive)" }
PaddleNLP
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment batch_size 16 Batch size learning_rate 1e-5 Learning rate train_path /opt/ml/input/data/dataset/train.txt Path of train file dev_path /opt/ml/input/data/dataset/dev.txt Path of dev file max_seq_len 512 Maximum sequence length num_epochs 100 Number of epochs seed 1000 seed logging_steps 10 Number of logging steps valid_steps 100 Number of validation steps device gpu CPU or GPU model uie-base Pretrained model name Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train dataset Content format of train/validation/test file
- Each row contains json string with content, result_list, and prompt, for example: { "content": "5月9日交通费29元从北苑到望京搜后", "result_list": [{"text": "5月9日", "start": 0, "end": 4}], "prompt": "时间" }
-
Quickstart - deploy
Environment variables
Environment variable Default value Comment device cpu CPU or GPU schema Schema to inference -
Quickstart - inference
HTTP request
{ "inputs":"上海虹桥高铁到杭州时间是9月24日费用是73元" }
HTTP response
{ "result": [ { "出发地": [ { "text": "上海", "start": 0, "end": 2, "probability": "0.99601215" } ], "目的地": [ { "text": "杭州", "start": 7, "end": 9, "probability": "0.99965054" } ], "费用": [ { "text": "73元", "start": 20, "end": 23, "probability": "0.79425305" } ], "时间": [ { "text": "9月24日", "start": 12, "end": 17, "probability": "0.9998573" } ] } ] }
DeBERTa
-
Create industrial model
Model extra information
{}
-
Quickstart - deploy
-
Quickstart - inference
HTTP request
{ "inputs": "如何有效学习?", "parameters": { "candidate_labels": [ "民生", "文化", "娱乐", "体育", "财经", "房产", "汽车", "教育", "科技", "军事", "旅游", "国际", "证券", "农业", "电竞" ] } }
HTTP response
{ "sequence": "如何有效学习?", "labels": [ "教育", "文化", "科技", "民生", "国际", "汽车", "军事", "电竞", "证券", "财经", "农业", "体育", "房产", "娱乐", "旅游" ], "scores": [ 0.7116793990135193, 0.03544081375002861, 0.0295342355966568, 0.028479689732193947, 0.024489011615514755, 0.023759257048368454, 0.023078029975295067, 0.017534229904413223, 0.01741044595837593, 0.017365973442792892, 0.01681639440357685, 0.01475503295660019, 0.013634421862661839, 0.013062535785138607, 0.012960496358573437 ] }
KeyBert
-
Create industrial model
Model extra information
{}
-
Quickstart - deploy
Environment variables
Environment variable Default value Comment type default One of [sentence-transformer, huggingface-transformer, flair, spacy, universal-sentence-encoder, gensim, default] model - paraphrase-multilingual-MiniLM-L12-v2 when type is sentence-transformer
- en_core_web_trf when type is Spacy
- https://tfhub.dev/google/universal-sentence-encoder/4 when type is universal-sentence-encoder
- fasttext-wiki-news-subwords-300 when type is gensimModel name huggingface_pipeline feature-extraction when type is huggingface-transformer HuggingFace pipeline mode - doc-embedding when type is flair
- transformer when type is spacy- Use document embeddeding model or word embedding model when type is flair
- Use non-transformer, transformer, transformer without GPU when type is spacydoc_embedding roberta-base when type is flair Document embedding word_embedding crawl when type is flair Word embedding keyphrase_ngram_start 1 Minimum length in words of generated keywords/keyphrases keyphrase_ngram_end 1 Maximum length in words of generated keywords/keyphrases stop_words english Stopwords to remove from the document top_n 5 Return the top n keywords/keyphrases min_df 1 Minimum document frequency of a word across all documents use_maxsum False Whether to use Max Sum Distance for the selection of keywords/keyphrases use_mmr False Whether to use Maximal Marginal Relevance (MMR) for the selection of keywords/keyphrases diversity 0.5 The diversity of the results between 0 and 1 if use_mmr
is set to Truenr_candidates 20 The number of candidates to consider if use_maxsum
is set to Truehighlight False Whether to print the document and highlight its keywords/keyphrases -
Quickstart - inference
HTTP request
{ "inputs": "近年来,嵌入式技术与无线网络技术深度结合,催生了可计算RFID、嵌入式传感网等新兴领域.这些系统由大量廉价的节点组成,应用前景广泛.在传统设计中,这些系统通常是根据应用定制的.根据应用定制的系统具有开发简便、运行高效等优点,但不适合未来大规模部署.这是因为如果这些系统跟应用密切绑定、难以更新,那么系统一经部署就难以更新其软件,从而阻碍了软件创新的进程.软件定义的思想可以有效解决该问题.当前,软件定义网络成为计算机网络中一个热门的研究领域.传感器网络的软件设计与因特网的软件设计存在诸多差异,其最大的差异在于,传感器网络主要以信息的采集为核心,而因特网主要以信息的传输为核心.此外,传感器节点还具有体积小、电池续航能力有限、价格低廉等特点.文中主要调研了设计软件定义传感器网络(Software-DefinedSensorNetworks,SDSNs)架构的相关工作,列举了在设计一个通用、高效的软件定义传感器网络架构时可能遇到的挑战,并回顾了一些有用的技术.这些技术有的来自于现有方案,有的能够直接被用来解决一部分挑战.此外,文中还从软件定义功能的角度,进一步地对目前通用、高效的软件定义传感器网络架构及其采用的技术进行了分类.我们认为,软件定义传感器网络架构将在已部署的网络中起到至关重要的作用,并带来一场新的技术变革." }
HTTP response
{ "result": [ [ "无线网络", 0.6223 ], [ "计算机网络", 0.447 ], [ "definedsensornetworks", 0.4356 ], [ "技术", 0.4297 ], [ "因特网", 0.4001 ] ] }
GluonTS
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment algo-name DeepAR Algorithm name model-dir 8 /opt/ml/model output-dir 20 /opt/ml/output train 0.001 /opt/ml/input/data/dataset test 0.9 /opt/ml/input/data/dataset freq 1D frequence prediction-length 2*14 Prediction length context-length 20*14 Context length batch-size 2048 Batch size epochs 100 Number of epochs num-batches-per-epoc 2 Number of batches per epochs learning-rate 0.001 Learning rate learning-rate-decay-factor 0.5 Learning rate decay factor patience 10 Patience minimum-learning-rate 5e-5 Minimum learning rate clip-gradient 10 Clip gradient weight-decay 1e-8 Weight decay init xavier hybridize False use-feat-dynamic-real False use-feat-static-cat False use-past-feat-dynamic-real False cardinality use-log1p False Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train data -
Quickstart - deploy
Environment variables
Environment variable Default value Comment freq 1H target_quantile 0.5 use_log1p False -
Quickstart - inference
HTTP request
{ "inputs": [ { "target": [ 12,28,12,0,56,12,8,12,28,4,24,8,12,4,4,8,16,4,24,12,16,16,4,24,12,12,16,24,8,16,12,40,12,32,24,8,8,8,20,20,32,20,16,16,8,24,32,12,16,4,4,28,12,24,20,12,44,32,36,20,24,28,40,84,52,20,20,64,36,44,28,24,32,12,44,72,50,15,35,80,75,24,60,42,12,18,60,54,72,49,72,88,24,88,90,33,22,9,54,88,117,136,99,80,50,110,88,44,44,36,63,72,108,63,60,55,66,33,120,55,12,100,80,117,72,135,55,77,120,66,121,132,72,0,60,77,50,136,40,88,99,77,275,168,26,80,110,130,140,130,187,110,204,220,144,204,60,36,156,220,154,279,297,374,132,192,182,266,140,156,247,228,360,270,209,104,247,364,420,255,240,225,300,338,352,429,273,336,270,390,570,544,765,208,336,405,434,364,416,256,368,493,408,666,1216,221,391,289,812,588,544,320,351,300,660,754,600,432,481,504,650,486,516,480,540,442,572,465,576,130,546,598,640,682,546,546,728,630,616,960,527,585,570,630,825,852,690,770,480,608,1072,1224,720,629,850,1037,988,767,935,900,1110,901,1560,2091,1008,833,1602,1785,1470,1365,1258,1152,1380,1332,2412,2320,1196,1350,1178,1136,117,363,846,595,1488,1691,2508,2992,725,1386,1701,1638,1548,1216,1360,1122,2180,1638,2814,2925,1363,902,1175,1320,1140,1292,1128,1134,1738 ], "start": "1980-01-01 00:00:00", "item_id": "T234", "freq": "1M", "prediction_length": 24 }, { "target": [ 12,28,12,0,56,12,8,12,28,4,24,8,12,4,4,8,16,4,24,12,16,16,4,24,12,12,16,24,8,16,12,40,12,32,24,8,8,8,20,20,32,20,16,16,8,24,32,12,16,4,4,28,12,24,20,12,44,32,36,20,24,28,40,84,52,20,20,64,36,44,28,24,32,12,44,72,50,15,35,80,75,24,60,42,12,18,60,54,72,49,72,88,24,88,90,33,22,9,54,88,117,136,99,80,50,110,88,44,44,36,63,72,108,63,60,55,66,33,120,55,12,100,80,117,72,135,55,77,120,66,121,132,72,0,60,77,50,136,40,88,99,77,275,168,26,80,110,130,140,130,187,110,204,220,144,204,60,36,156,220,154,279,297,374,132,192,182,266,140,156,247,228,360,270,209,104,247,364,420,255,240,225,300,338,352,429,273,336,270,390,570,544,765,208,336,405,434,364,416,256,368,493,408,666,1216,221,391,289,812,588,544,320,351,300,660,754,600,432,481,504,650,486,516,480,540,442,572,465,576,130,546,598,640,682,546,546,728,630,616,960,527,585,570,630,825,852,690,770,480,608,1072,1224,720,629,850,1037,988,767,935,900,1110,901,1560,2091,1008,833,1602,1785,1470,1365,1258,1152,1380,1332,2412,2320,1196,1350,1178,1136,117,363,846,595,1488,1691,2508,2992,725,1386,1701,1638,1548,1216,1360,1122,2180,1638,2814,2925,1363,902,1175,1320,1140,1292,1128,1134,1738,1564,3427,3325,1482,1584,1449,2222,1836,1332,1584,1680,1869,2002,3190,2652,1568,966,1380,1794,1350,1577,1104,1188,1495 ], "start":"1980-01-01 00:00:00", "item_id": "T234", "freq": "1M", "prediction_length": 24 } ] }
HTTP response
{ "result": [ [ 1930.8326416015625, 2290.176025390625, 2243.770263671875, 1112.35888671875, 1216.8353271484375, 1300.52587890625, 1319.37451171875, 1104.0013427734375, 1231.51611328125, 1281.2568359375, 1306.0784912109375, 1798.2593994140625, 1773.2158203125, 2148.71142578125, 2055.391357421875, 1097.5042724609375, 1143.5330810546875, 1365.8883056640625, 1335.6077880859375, 1208.5709228515625, 1230.0604248046875, 1293.724853515625, 1298.18798828125, 1691.08056640625 ], [ 1820.448974609375, 2697.013916015625, 2449.663818359375, 1375.6854248046875, 1132.6513671875, 1368.69091796875, 1617.199951171875, 1349.854248046875, 1284.10986328125, 1291.0433349609375, 1288.84130859375, 1595.6607666015625, 1741.90185546875, 2308.14013671875, 2105.67529296875, 1253.835205078125, 1139.4034423828125, 1357.335205078125, 1589.59228515625, 1372.781982421875, 1298.3154296875, 1244.7462158203125, 1267.4632568359375, 1574.6981201171875 ] ] }
StyleGAN
-
Create industrial model
Model extra information
{}
-
Quickstart - train
Hyperparameters
Hyperparameter Default value Comment gpus 1 Num of GPUS snap 50 Number of GPUs to use [default: 1] metrics fid50k_full Comma-separated list or "none" [default: fid50k_full] seed 0 Random seed data Training data (directory or zip) cond false Train conditional model based on dataset labels [default: false] subset all 'Train with only N images mirror false Enable dataset x-flips config auto Base config, one of 'auto', 'stylegan2', 'paper256', 'paper512', 'paper1024', 'cifar' gamma Override R1 gamma kimg Override training duration batch Override batch size aug ada Augmentation mode, one of 'noaug', 'ada', 'fixed' p Augmentation probability for --aug=fixed target ADA target value for --aug=ada augpipe bgc Augmentation pipeline resume noresume Resume training freezed 0 Freezed layers fp32 Disable mixed-precision training nhwc Use NHWC memory format with FP16 nobench Disable cuDNN benchmarking allow-tf32 Allow PyTorch to use TF32 internally workers Override number of DataLoader workers Input data configuration
Channel name Mandatory Comment dataset Yes S3 URI of train data -
Quickstart - deploy
Environment variables
Environment variable Default value Comment network Pre-trianed network URL -
Quickstart - inference
HTTP request
{ "inputs": { "trunc": "0.7325021066422363", "seeds": "9,206,870,370" } }
HTTP response
[ "s3://sagemaker-ap-east-1-034068151705/stylegan/inference/outputseed0370.png", "s3://sagemaker-ap-east-1-034068151705/stylegan/inference/outputseed0009.png", "s3://sagemaker-ap-east-1-034068151705/stylegan/inference/outputseed0870.png", "s3://sagemaker-ap-east-1-034068151705/stylegan/inference/outputseed0206.png" ]
Yolov5PaddleOCR
-
Create industrial model
Model extra information
{}
-
Quickstart - inference
stable-diffusion-webui
-
Create industrial model
It will be created by default if you have started stable-diffusion-webui once. Alternative you can create it explicitly. Note industrial model of stable-diffusion-webui is unique within one all-in-one-ai app and with name 'stable-diffusion-webui' by design.
-
Quickstart - train
Basically we support 3 train approach instable-diffusion-webui: embedding, hypernetwork, and dreambooth which can be used to train person, object, style.
Strongly recommend that you start the training job inside of stable-diffusion-webui since it is already supported with more friendely user interface.
Alternative you start the training job explicitly.
Hyperparameters
Hyperparameter Default value Comment region Current region name AWS region name embeddings-s3uri s3://[sagemaker-default-bucket]/stable-diffusion-webui/embeddings/ S3 URI of embeddings, only applicable for embedding or hypernetwork training hypernetwork-s3uri s3://[sagemaker-default-bucket]/stable-diffusion-webui/hypernetwork/ S3 URI of hypernetwork, only applicable for embedding or hypernetwork training train-task embedding One of embedding, hypernetwork, dreambooth api-endpoint REST API Gateway of all-in-one-ai REST API Gateway db-models-s3uri s3://[sagemaker-default-bucket]/stable-diffusion-webui/dreambooth/ S3 URI of dreambooth model S3 URI, only applicable for dreambooth training sd-models-s3uri s3://[sagemaker-default-bucket]/stable-diffusion-webui/models/ stable diffusion models S3 URI, only applicable for dreambooth training train-args train-args which is up to train-task dreambooth-config-id dreambooth config id which is used to identify the dreambooth config in s3://[sagemaker-default-bucket]/stable-diffusion-webui/dreambooth-config/ train-args example for train dreambooth
{\"train_dreambooth_settings\": {\"db_create_new_db_model\": true, \"db_new_model_name\": \"my-awsdogtoy-model-002\", \"db_new_model_src\": \"768-v-ema.ckpt\", \"db_new_model_scheduler\": \"ddim\", \"db_create_from_hub\": false, \"db_new_model_url\": \"\", \"db_new_model_token\": \"\", \"db_new_model_extract_ema\": false, \"db_model_name\": \"\", \"db_lora_model_name\": \"\", \"db_lora_weight\": 1, \"db_lora_txt_weight\": 1, \"db_train_imagic_only\": false, \"db_use_subdir\": false, \"db_custom_model_name\": \"\", \"db_train_wizard_person\": false, \"db_train_wizard_object\": true, \"db_performance_wizard\": true}}
dreambooth-config example for train dreambooth
""" model_name: str = "", adam_beta1: float = 0.9, adam_beta2: float = 0.999, adam_epsilon: float = 1e-8, adam_weight_decay: float = 0.01, attention: str = "default", center_crop: bool = True, concepts_path: str = "", custom_model_name: str = "", epoch_pause_frequency: int = 0, epoch_pause_time: int = 0, gradient_accumulation_steps: int = 1, gradient_checkpointing: bool = True, half_model: bool = False, has_ema: bool = False, hflip: bool = False, learning_rate: float = 0.00000172, lora_learning_rate: float = 1e-4, lora_txt_learning_rate: float = 5e-5, lr_scheduler: str = 'constant', lr_warmup_steps: int = 0, max_token_length: int = 75, max_train_steps: int = 1000, mixed_precision: str = "fp16", model_path: str = "", not_cache_latents=False, num_train_epochs: int = 1, pad_tokens: bool = True, pretrained_vae_name_or_path: str = "", prior_loss_weight: float = 1.0, resolution: int = 512, revision: int = 0, sample_batch_size: int = 1, save_class_txt: bool = False, save_embedding_every: int = 500, save_preview_every: int = 500, save_use_global_counts: bool = False, save_use_epochs: bool = False, scale_lr: bool = False, scheduler: str = "ddim", src: str = "", shuffle_tags: bool = False, train_batch_size: int = 1, train_text_encoder: bool = True, use_8bit_adam: bool = True, use_concepts: bool = False, use_cpu: bool = False, use_ema: bool = True, use_lora: bool = False, v2: bool = False, c1_class_data_dir: str = "", c1_class_guidance_scale: float = 7.5, c1_class_infer_steps: int = 60, c1_class_negative_prompt: str = "", c1_class_prompt: str = "", c1_class_token: str = "", c1_instance_data_dir: str = "", c1_instance_prompt: str = "", c1_instance_token: str = "", c1_max_steps: int = -1, c1_n_save_sample: int = 1, c1_num_class_images: int = 0, c1_sample_seed: int = -1, c1_save_guidance_scale: float = 7.5, c1_save_infer_steps: int = 60, c1_save_sample_negative_prompt: str = "", c1_save_sample_prompt: str = "", c1_save_sample_template: str = "", c2_class_data_dir: str = "", c2_class_guidance_scale: float = 7.5, c2_class_infer_steps: int = 60, c2_class_negative_prompt: str = "", c2_class_prompt: str = "", c2_class_token: str = "", c2_instance_data_dir: str = "", c2_instance_prompt: str = "", c2_instance_token: str = "", c2_max_steps: int = -1, c2_n_save_sample: int = 1, c2_num_class_images: int = 0, c2_sample_seed: int = -1, c2_save_guidance_scale: float = 7.5, c2_save_infer_steps: int = 60, c2_save_sample_negative_prompt: str = "", c2_save_sample_prompt: str = "", c2_save_sample_template: str = "", c3_class_data_dir: str = "", c3_class_guidance_scale: float = 7.5, c3_class_infer_steps: int = 60, c3_class_negative_prompt: str = "", c3_class_prompt: str = "", c3_class_token: str = "", c3_instance_data_dir: str = "", c3_instance_prompt: str = "", c3_instance_token: str = "", c3_max_steps: int = -1, c3_n_save_sample: int = 1, c3_num_class_images: int = 0, c3_sample_seed: int = -1, c3_save_guidance_scale: float = 7.5, c3_save_infer_steps: int = 60, c3_save_sample_negative_prompt: str = "", c3_save_sample_prompt: str = "", c3_save_sample_template: str = "", concepts_list=None """ [ "", 0.9, 0.999, 1e-08, 0.01, "default", False, "", "", 0.0, 60.0, 1, True, False, "", True, 2e-06, 0.0002, 0.0002, "constant", 500, 75, 0, "no", "", True, 100, True, "", 1, 512, "", 1, True, 500, 500, True, False, False, "", "", False, 1, True, False, False, False, False, False, "", "", 7.5, 40, "", "", "photo of dog", "/opt/ml/input/data/concepts/images", "", "photo of awsdogtoy dog", -1, 1, 0, -1, 7.5, 40, "", "", "", "", 7.5, 40, "", "", "", "", "", "", -1, 1, 0, -1, 7.5, 40, "", "", "", "", 7.5, 40, "", "", "", "", "", "", -1, 1, 0, -1, 7.5, 40, "", "", "" ]
Input data configuration
Channel name Mandatory Comment images Yes S3 URI of images models No S3 URI of stable diffusion models embedding No S3 URI of embeddings hypernetwork No S3 URI of hypernetwork lora No S3 URI of lora models dreambooth No S3 URI of dreambooth models -
Quickstart - deploy
Environment variables
Environment variable Default value Comment api_endpoint REST API Gateway of all-in-one-ai REST API Gateway endpoint_name Name of SageMaker Endpoint which is be used to host stable diffusion models and generate images -
Quickstart - Inference - Text to Image
HTTP request
payload = { 'enable_hr': False, 'denoising_strength': 0.7, 'firstphase_width': 0, 'firstphase_height': 0, 'prompt': "dog", 'styles': ['None', 'None'], 'seed': -1, 'subseed': -1, 'subseed_strength': 0.0, 'seed_resize_from_h': 0, 'seed_resize_from_w': 0, 'sampler_name': None, 'batch_size': 1, 'n_iter':1, 'steps': 20, 'cfg_scale': 7.0, 'width': 768, 'height': 768, 'restore_faces': False, 'tiling': False, 'negative_prompt': '', 'eta': 1.0, 's_churn': 0.0, 's_tmax': None, 's_tmin': 0.0, 's_noise': 1.0, 'override_settings': {}, 'script_args': '[0, false, false, false, "", 1, "", 0, "", true, false, false]', 'sampler_index': 'Euler a' } inputs = { 'task': 'text-to-image', 'txt2img_payload': payload, 'username': 'e' }
HTTP response
{ "images" : [ [base64 encoded images], ..., [base64 encoded images] ] }
-
Quickstart - Inference - Image to Image
HTTP request
payload = { 'init_images': [image_encoded_in_base64], 'resize_mode': 0, 'denoising_strength': 0.75, 'mask': None, 'mask_blur': 4, 'inpainting_fill': 1, 'inpaint_full_res': False, 'inpaint_full_res_padding': 32, 'inpainting_mask_invert': 0, 'prompt': 'cat', 'styles': ['None', 'None'], 'seed': -1, 'subseed': -1, 'subseed_strength': 0.0, 'seed_resize_from_h': 0, 'seed_resize_from_w': 0, 'sampler_name': None, 'batch_size': 1, 'n_iter': 1, 'steps': 20, 'cfg_scale': 7.0, 'width': 768, 'height': 768, 'restore_faces': False, 'tiling': False, 'negative_prompt': '', 'eta': 1.0, 's_churn': 0.0, 's_tmax': None, 's_tmin': 0.0, 's_noise': 1.0, 'override_settings': {}, 'script_args': '[0, "<ul>\\n<li><code>CFG Scale</code> should be 2 or lower.</li>\\n</ul>\\n", true, true, "", "", true, 50, true, 1, 0, false, 4, 1, "<p style=\\"margin-bottom:0.75em\\">Recommended settings: Sampling Steps: 80-100, Sampler: Euler a, Denoising strength: 0.8</p>", 128, 8, ["left", "right", "up", "down"], 1, 0.05, 128, 4, 0, ["left", "right", "up", "down"], false, false, false, "", "<p style=\\"margin-bottom:0.75em\\">Will upscale the image to twice the dimensions; use width and height sliders to set tile size</p>", 64, 0, 1, "", 0, "", true, false, false]', 'sampler_index': 'Euler a', 'include_init_images': False } inputs = { 'task': 'image-to-image', 'img2img_payload': payload, 'username': 'e' }
HTTP response
{ "images" : [ [base64 encoded images], ..., [base64 encoded images] ] }