From 02403feb54d6f4bee60628d41f3412afd57d2c95 Mon Sep 17 00:00:00 2001 From: mozhou52 Date: Mon, 29 Sep 2025 14:20:03 +0800 Subject: [PATCH] feat: add e2e model tests and refactor model deployment logic - Add model download and deployment tests in CI script - Refactor NAS and OSS mount point configurations in function deployment - Remove outdated example files and test scripts - Simplify model server address generation logic - Update model_test.ts to match the correct endpoint for DevClient Co-authored-by: Qwen-Coder --- __tests__/e2e/ci-mac-linux.sh | 3 + __tests__/e2e/model/deploy_and_test_model.py | 42 ++- package-lock.json | 14 +- package.json | 2 +- src/subCommands/model/index.ts | 262 +++++++++++-------- 5 files changed, 195 insertions(+), 128 deletions(-) diff --git a/__tests__/e2e/ci-mac-linux.sh b/__tests__/e2e/ci-mac-linux.sh index cee46330..e595b7a9 100755 --- a/__tests__/e2e/ci-mac-linux.sh +++ b/__tests__/e2e/ci-mac-linux.sh @@ -31,8 +31,11 @@ echo "test model download" cd model pip install -r requirements.txt export fc_component_function_name=model-$(uname)-$(uname -m)-$RANDSTR +# export artifact_endpoint=devs-pre.cn-hangzhou.aliyuncs.com python deploy_and_test_model.py --model-id iic/cv_LightweightEdge_ocr-recognitoin-general_damo --region cn-shanghai --auto-cleanup python deploy_and_test_model.py --model-id Qwen/Qwen2.5-0.5B-Instruct --region cn-shanghai --auto-cleanup +# python deploy_and_test_model.py --model-id iic/cv_LightweightEdge_ocr-recognitoin-general_damo --region cn-shanghai --storage oss --auto-cleanup +# python deploy_and_test_model.py --model-id Qwen/Qwen2.5-0.5B-Instruct --region cn-shanghai --storage oss --auto-cleanup cd .. echo "test go runtime" diff --git a/__tests__/e2e/model/deploy_and_test_model.py b/__tests__/e2e/model/deploy_and_test_model.py index f7e51d82..19e40492 100644 --- a/__tests__/e2e/model/deploy_and_test_model.py +++ b/__tests__/e2e/model/deploy_and_test_model.py @@ -42,7 +42,7 @@ def simple_hash(input_string: str) -> str: return sha256_hash[:8] + sha256_hash[-8:] -def deploy_model(model_id: str, region: str = "cn-hangzhou"): +def deploy_model(model_id: str, region: str = "cn-hangzhou", storage: str = "nas"): """ 部署模型到函数计算 @@ -71,9 +71,7 @@ def deploy_model(model_id: str, region: str = "cn-hangzhou"): encoded_model_id = urllib.parse.quote(model_id, safe="") # 调用部署接口 - model_registry_url = os.getenv( - "MODEL_REGISTRY_URL", "model-registry.devsapp.cn" - ) + model_registry_url = os.getenv("MODEL_REGISTRY_URL", "model-registry.devsapp.cn") deploy_url = ( f"http://{model_registry_url}/api/v1/models/{encoded_model_id}/deploy-info" ) @@ -108,6 +106,28 @@ def deploy_model(model_id: str, region: str = "cn-hangzhou"): # 更新resources中的props s_yaml["resources"]["test_func"]["props"] = deploy_info + # 下载到 oss + if storage == "oss": + s_yaml["resources"]["test_func"]["props"].pop("nasConfig", None) + s_yaml["resources"]["test_func"]["props"].pop("vpcConfig", None) + s_yaml["resources"]["test_func"]["props"]["ossMountConfig"] = "auto" + s_yaml["resources"]["test_func"]["props"][ + "role" + ] = "acs:ram::${config('AccountID')}:role/aliyunfcdefaultrole" + s_yaml["resources"]["test_func"]["props"]["annotations"]["modelConfig"][ + "storage" + ] = storage + # 修改 customContainerConfig.entrypoint 中的路径 + custom_container_config = s_yaml["resources"]["test_func"]["props"].get("customContainerConfig", {}) + if "entrypoint" in custom_container_config: + entrypoint = custom_container_config["entrypoint"] + if isinstance(entrypoint, list): + # 遍历 entrypoint 数组，替换包含 /mnt/ 的路径 + for i, item in enumerate(entrypoint): + if isinstance(item, str) and "/mnt/" in item and not item.startswith("vllm") and not item.isdigit() and item not in ["--port", "--served-model-name", "--trust-remote-code"]: + entrypoint[i] = f"/mnt/serverless-{region}-d5d4cd07-616a-5428-91b7-ec2d0257b3a2" + custom_container_config["entrypoint"] = entrypoint + # 保存配置到临时文件 s_yaml_file = f"s.yaml" with open(s_yaml_file, "w", encoding="utf-8") as f: @@ -326,14 +346,9 @@ def cleanup_deployment(s_yaml_file: str): try: print(f"正在清除部署资源: {s_yaml_file}") # 清除模型 - subprocess.check_call( - f"echo 123456 | sudo -S s model remove -y -t {s_yaml_file}", shell=True - ) + subprocess.check_call(f"s model remove -y -t {s_yaml_file}", shell=True) # 清除函数 - subprocess.check_call( - f"echo 123456 | sudo -S s remove -y -t {s_yaml_file} --skip-push", - shell=True, - ) + subprocess.check_call(f"s remove -y -t {s_yaml_file} --skip-push", shell=True) print("部署资源清除完成!") except subprocess.CalledProcessError as e: print(f"清除部署资源失败: {e}") @@ -354,6 +369,7 @@ def main(): parser.add_argument( "--auto-cleanup", action="store_true", help="部署和测试完成后自动执行清理操作" ) + parser.add_argument("--storage", help="存储区域", default="nas") args = parser.parse_args() @@ -364,7 +380,9 @@ def main(): return 0 elif args.model_id: # 部署和测试模型 - deploy_url, s_yaml_file = deploy_model(args.model_id, args.region) + deploy_url, s_yaml_file = deploy_model( + args.model_id, args.region, args.storage + ) # 测试模型 test_model(args.model_id, deploy_url, s_yaml_file) diff --git a/package-lock.json b/package-lock.json index 5a20f089..392eaa0f 100644 --- a/package-lock.json +++ b/package-lock.json @@ -10,7 +10,7 @@ "hasInstallScript": true, "license": "ISC", "dependencies": { - "@alicloud/devs20230714": "^2.4.5", + "@alicloud/devs20230714": "^2.4.6-alpha.2", "@alicloud/fc2": "^2.6.6", "@alicloud/fc20230330": "4.6.0", "@alicloud/pop-core": "^1.8.0", @@ -154,9 +154,9 @@ } }, "node_modules/@alicloud/devs20230714": { - "version": "2.4.5", - "resolved": "https://packages.aliyun.com/670e108663cd360abfe4be65/npm/npm-registry/@alicloud/devs20230714/-/@alicloud/devs20230714-2.4.5.tgz", - "integrity": "sha512-4gBq5S014v9jICo/asDbpcXIULaR4aqjB4E4IqDvweIJzmMXvENkHeZbklZNqzep5OmTUCJPFNqJ1dVhohjHsQ==", + "version": "2.4.6-alpha.2", + "resolved": "https://packages.aliyun.com/670e108663cd360abfe4be65/npm/npm-registry/@alicloud/devs20230714/-/@alicloud/devs20230714-2.4.6-alpha.2.tgz", + "integrity": "sha512-7gSRncwItnPWNDcs91PqILLsZCcfeZ/C8iwbo/6Z3S8mNMIvtxXYyGjwOWncHCOqVxQOvt6vzIEH/0KXeNyOHg==", "license": "Apache-2.0", "dependencies": { "@alicloud/openapi-core": "^1.0.0", @@ -16113,9 +16113,9 @@ } }, "@alicloud/devs20230714": { - "version": "2.4.5", - "resolved": "https://packages.aliyun.com/670e108663cd360abfe4be65/npm/npm-registry/@alicloud/devs20230714/-/@alicloud/devs20230714-2.4.5.tgz", - "integrity": "sha512-4gBq5S014v9jICo/asDbpcXIULaR4aqjB4E4IqDvweIJzmMXvENkHeZbklZNqzep5OmTUCJPFNqJ1dVhohjHsQ==", + "version": "2.4.6-alpha.2", + "resolved": "https://packages.aliyun.com/670e108663cd360abfe4be65/npm/npm-registry/@alicloud/devs20230714/-/@alicloud/devs20230714-2.4.6-alpha.2.tgz", + "integrity": "sha512-7gSRncwItnPWNDcs91PqILLsZCcfeZ/C8iwbo/6Z3S8mNMIvtxXYyGjwOWncHCOqVxQOvt6vzIEH/0KXeNyOHg==", "requires": { "@alicloud/openapi-core": "^1.0.0", "@darabonba/typescript": "^1.0.0" diff --git a/package.json b/package.json index 2d60461c..759a4166 100644 --- a/package.json +++ b/package.json @@ -22,7 +22,7 @@ "author": "", "license": "ISC", "dependencies": { - "@alicloud/devs20230714": "^2.4.5", + "@alicloud/devs20230714": "^2.4.6-alpha.2", "@alicloud/fc2": "^2.6.6", "@alicloud/fc20230330": "4.6.0", "@alicloud/pop-core": "^1.8.0", diff --git a/src/subCommands/model/index.ts b/src/subCommands/model/index.ts index 56bc9b11..d1b1783d 100644 --- a/src/subCommands/model/index.ts +++ b/src/subCommands/model/index.ts @@ -5,13 +5,15 @@ import FC from '../../resources/fc'; import VPC_NAS from '../../resources/vpc-nas'; import { ICredentials } from '@serverless-devs/component-interface'; import { yellow } from 'chalk'; -import DevClient, * as devs from '@alicloud/devs20230714'; +import DevClient, { DownloadModelRequest } from '@alicloud/devs20230714'; import * as $OpenApi from '@alicloud/openapi-client'; import { getEnvVariable } from '../../default/resources'; import commandsHelp from '../../commands-help/layer'; import { parseArgv } from '@serverless-devs/utils'; import assert from 'assert'; import { sleep } from '../../utils'; +import OSS from '../../resources/oss'; +import { OSSMountPoint, VPCConfig } from '@alicloud/fc20230330'; export const NEW_MODEL_SERVICE_CLIENT_CONNECT_TIMEOUT: number = parseInt(process.env.NEW_MODEL_SERVICE_CLIENT_CONNECT_TIMEOUT as string, 10) || 60 * 1000; @@ -76,9 +78,40 @@ export class Model { this.envName = envName; const name = `${projectName}$${envName}$${functionName}`; - const { nasAuto, vpcAuto } = FC.computeLocalAuto(this.local); - logger.debug(`[nasAuto] Auto compute local auto, nasAuto: ${nasAuto};`); - if (nasAuto || vpcAuto) { + const { nasAuto, vpcAuto, ossAuto } = FC.computeLocalAuto(this.local); + logger.debug(`[auto] Auto compute local auto, nasAuto: ${nasAuto} ossAuto: ${ossAuto};`); + + if (modelConfig?.storage === 'oss' && ossAuto) { + let ossEndpoint = `https://oss-${region}.aliyuncs.com`; + if (process.env.FC_REGION === region) { + ossEndpoint = `oss-${region}-internal.aliyuncs.com`; + } + logger.info(`ossAuto code to ${ossEndpoint}`); + const oss = new OSS(region, credential as ICredentials, ossEndpoint); + const { ossBucket } = await oss.deploy(); + logger.write( + yellow(`Created oss resource succeeded, please replace ossMountConfig: auto in yaml with: +ossMountConfig: + mountPoints: + - mountDir: /mnt/${ossBucket} + bucketName: ${ossBucket} + endpoint: http://oss-${region}-internal.aliyuncs.com + readOnly: false\n`), + ); + this.createResource.oss = { ossBucket }; + _.set(this.local, 'ossMountConfig', { + mountPoints: [ + { + mountDir: `/mnt/${ossBucket}`, + bucketName: ossBucket, + endpoint: `http://oss-${region}-internal.aliyuncs.com`, + readOnly: false, + }, + ], + }); + } + + if (modelConfig.storage === 'nas' && (nasAuto || vpcAuto)) { const client = new VPC_NAS(region, credential as ICredentials); const localVpcAuto = _.isString(this.local.vpcConfig) ? undefined : this.local.vpcConfig; // @ts-ignore: nas auto 会返回 mountTargetDomain 和 fileSystemId @@ -137,125 +170,138 @@ mountPoints: const devClient = await this.getNewModelServiceClient(); if (devClient) { - const { nasConfig, vpcConfig } = this.local; - logger.debug(`[Download-model] nasConfig: ${nasConfig} vpcConfig: ${vpcConfig}`); - const { nasMountDomain, nasMountPath } = this.parseNasConfig(nasConfig); + const { nasConfig, vpcConfig, ossMountConfig } = this.local; + logger.debug( + `[Download-model] nasConfig: ${nasConfig} vpcConfig: ${vpcConfig} ossMountConfig: ${ossMountConfig}`, + ); let resp; - if (typeof nasConfig === 'object' && nasConfig?.mountPoints) { - const params = { - modelConfig: { - model: modelConfig.id, - type: modelConfig.source, - reversion: modelConfig.version, - token: modelConfig.token, - bucket: modelConfig.ossBucket, - path: modelConfig.ossPath, - region: modelConfig.ossRegion, - }, - region, - modelPath: nasMountPath, - role: modelConfig.role, - nasMountPoint: nasMountDomain, - vpcConfig, - syncStrategy: process.env.MODEL_DOWNLOAD_STRATEGY || 'incremental_once', - }; - try { - const req = new devs.DownloadModelRequest(params); - logger.debug(req); - resp = await devClient.downloadModel(name, req); - logger.debug(resp); - } catch (e) { - logger.error(`download model invocation error: ${e.message}`); - throw new Error(`download model error: ${e.message}`); - } + const params: any = { + modelConfig: { + model: modelConfig.id, + type: modelConfig.source, + reversion: modelConfig.version, + token: modelConfig.token, + bucket: modelConfig.ossBucket, + path: modelConfig.ossPath, + region: modelConfig.ossRegion, + }, + region, + role: modelConfig.role, + syncStrategy: process.env.MODEL_DOWNLOAD_STRATEGY || 'incremental_once', + }; + + if ( + modelConfig.storage === 'oss' && + typeof ossMountConfig === 'object' && + ossMountConfig?.mountPoints + ) { + params.ossMountPoints = ossMountConfig.mountPoints as OSSMountPoint[]; + } else if ( + modelConfig.storage === 'nas' && + typeof nasConfig === 'object' && + nasConfig?.mountPoints + ) { + const { nasMountDomain, nasMountPath } = this.parseNasConfig(nasConfig); + params.vpcConfig = vpcConfig as VPCConfig; + params.nasMountPoint = nasMountDomain; + params.modelPath = nasMountPath; + } + try { + const req = new DownloadModelRequest(params); + logger.debug(req); + resp = await devClient.downloadModel(name, req); + logger.debug(resp); + } catch (e) { + logger.error(`download model invocation error: ${e.message}`); + throw new Error(`download model error: ${e.message}`); + } - if (resp.statusCode !== 200 && resp.statusCode !== 202) { - logger.info({ status: resp.statusCode, body: resp.body }); - throw new Error( - `download model connection error, statusCode: ${resp.statusCode}, body: ${resp.body}`, - ); - } + if (resp.statusCode !== 200 && resp.statusCode !== 202) { + logger.info({ status: resp.statusCode, body: resp.body }); + throw new Error( + `download model connection error, statusCode: ${resp.statusCode}, body: ${resp.body}`, + ); + } - const rb = resp.body; - if (rb.success || rb.errMsg.includes('is already exist')) { - logger.info(`download model requestId: ${rb.requestId}`); - const shouldContinue = true; - while (shouldContinue) { - // eslint-disable-next-line no-await-in-loop - const modelStatus = await this.getModelStatus(devClient, name); - - if (modelStatus.finished) { - if ( - modelStatus.total && - modelStatus.currentBytes !== undefined && - modelStatus.fileSize !== undefined - ) { - const currentMB = (modelStatus.currentBytes / 1024 / 1024).toFixed(1); - const totalMB = (modelStatus.fileSize / 1024 / 1024).toFixed(1); - - const totalBars = 50; - const progressBar = '='.repeat(totalBars); - - process.stdout.write( - `\r[Download-model] [${progressBar}] 100.00% (${currentMB}MB/${totalMB}MB)\n`, - ); - } else { - process.stdout.write('\n'); - } - // 清除进度条并换行 - process.stdout.write('\n'); - if (modelStatus.total) { - const durationMs = modelStatus.finishedTime - modelStatus.startTime; - const durationSeconds = Math.floor(durationMs / 1000); - logger.info(`Time taken for model download: ${durationSeconds}s.`); - } - logger.info(`[Download-model] Download model finished.`); - return true; - } - // 显示下载进度 - if (modelStatus.currentBytes !== undefined && modelStatus.fileSize !== undefined) { - const percentage = (modelStatus.currentBytes / modelStatus.fileSize) * 100; + const rb = resp.body; + if (rb.success || rb.errMsg.includes('is already exist')) { + logger.info(`download model requestId: ${rb.requestId}`); + const shouldContinue = true; + while (shouldContinue) { + // eslint-disable-next-line no-await-in-loop + const modelStatus = await this.getModelStatus(devClient, name); + + if (modelStatus.finished) { + if ( + modelStatus.total && + modelStatus.currentBytes !== undefined && + modelStatus.fileSize !== undefined + ) { const currentMB = (modelStatus.currentBytes / 1024 / 1024).toFixed(1); const totalMB = (modelStatus.fileSize / 1024 / 1024).toFixed(1); - // 每个等号代表2%，向下取整计算等号数量 - const totalBars = 50; // 总共50个字符位置 - const filledBars = Math.min(totalBars, Math.floor(percentage / 2)); // 每个等号代表2% - const emptyBars = totalBars - filledBars; - - const progressBar = '='.repeat(filledBars) + '.'.repeat(emptyBars); + const totalBars = 50; + const progressBar = '='.repeat(totalBars); process.stdout.write( - `\r[Download-model] [${progressBar}] ${percentage.toFixed( - 2, - )}% (${currentMB}MB/${totalMB}MB)`, + `\r[Download-model] [${progressBar}] 100.00% (${currentMB}MB/${totalMB}MB)\n`, ); - } - - if (Date.now() - modelStatus.startTime > MODEL_DOWNLOAD_TIMEOUT) { - // 清除进度条并换行 + } else { process.stdout.write('\n'); - const errorMessage = `[Model-download] Download timeout after ${ - MODEL_DOWNLOAD_TIMEOUT / 1000 / 60 - } minutes`; - throw new Error(errorMessage); } - - // 根据文件大小调整轮询间隔 - let sleepTime = 2; // 默认2秒 - if (modelStatus.fileSize !== undefined && modelStatus.fileSize > 1024 * 1024 * 1024) { - // 文件大于1GB时，轮询间隔为10秒 - sleepTime = 10; + // 清除进度条并换行 + process.stdout.write('\n'); + if (modelStatus.total) { + const durationMs = modelStatus.finishedTime - modelStatus.startTime; + const durationSeconds = Math.floor(durationMs / 1000); + logger.info(`Time taken for model download: ${durationSeconds}s.`); } + logger.info(`[Download-model] Download model finished.`); + return true; + } + // 显示下载进度 + if (modelStatus.currentBytes !== undefined && modelStatus.fileSize !== undefined) { + const percentage = (modelStatus.currentBytes / modelStatus.fileSize) * 100; + const currentMB = (modelStatus.currentBytes / 1024 / 1024).toFixed(1); + const totalMB = (modelStatus.fileSize / 1024 / 1024).toFixed(1); + + // 每个等号代表2%，向下取整计算等号数量 + const totalBars = 50; // 总共50个字符位置 + const filledBars = Math.min(totalBars, Math.floor(percentage / 2)); // 每个等号代表2% + const emptyBars = totalBars - filledBars; + + const progressBar = '='.repeat(filledBars) + '.'.repeat(emptyBars); + + process.stdout.write( + `\r[Download-model] [${progressBar}] ${percentage.toFixed( + 2, + )}% (${currentMB}MB/${totalMB}MB)`, + ); + } + + if (Date.now() - modelStatus.startTime > MODEL_DOWNLOAD_TIMEOUT) { + // 清除进度条并换行 + process.stdout.write('\n'); + const errorMessage = `[Model-download] Download timeout after ${ + MODEL_DOWNLOAD_TIMEOUT / 1000 / 60 + } minutes`; + throw new Error(errorMessage); + } - // eslint-disable-next-line no-await-in-loop - await sleep(sleepTime); + // 根据文件大小调整轮询间隔 + let sleepTime = 2; // 默认2秒 + if (modelStatus.fileSize !== undefined && modelStatus.fileSize > 1024 * 1024 * 1024) { + // 文件大于1GB时，轮询间隔为10秒 + sleepTime = 10; } - } else { - throw new Error( - `download model service biz failed, errCode: ${rb.errCode}, errMsg: ${rb.errMsg}`, - ); + + // eslint-disable-next-line no-await-in-loop + await sleep(sleepTime); } + } else { + throw new Error( + `download model service biz failed, errCode: ${rb.errCode}, errMsg: ${rb.errMsg}`, + ); } } }