mitzen

While current llm model mostly uses safe tensor, we can force an existing model to save and load safe tensor by using the following command  ! pip install torch ! pip install -U transformers datasets evaluate accelerate timm from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained( "Qwen/Qwen3.5-2B" , dtype= "auto" , device_map= "auto" ) tokenizer = AutoTokenizer.from_pretrained( "Qwen/Qwen3.5-2B" ) model_inputs = tokenizer([ "The secret to baking a good cake is " ], return_tensors= "pt" ).to(model.device) generated_ids = model.generate(**model_inputs, max_length= 30 ) tokenizer.batch_decode(generated_ids)[ 0 ] ## This is where we save the model as safe tensor model.save_pretrained( "model" , safe_serialization= True ) Then we can reload it using this code model_safe = AutoModelForCausalLM.from_pretrained(     "./model" ,     trust_remote_code= True ) gener...

First we will create AWS lambda and then configure our app to point to it via Gateway API.  Create AWS lambda using the following steps:- Sign in to the Lambda console at https://console.aws.amazon.com/lambda. 1.Choose Create function. For Function name, enter my-function. 2. For all other options, use the default setting. Choose Create function. Gateway API Configuring the API gateway is abit tricky. Goto https://console.aws.amazon.com/apigateway. And the click on "Create API".   Then select HTTP API and click on "build". Provide a name, in this case we will call it "my-http-api" and add the required integration to lambda as shown here:- Then in the route settings, ensure you have setup the followings In the define "stage", click "Next" and please verify the configuration is ok. Click "Create". And then to test it out, go to API Gateway -> APIS -> "my-http-api". Then go to Deploy -> Stage and click on the I...

Ran into this error when trying to deploy agent to vertex ai agent runtime.  Base on the tutorial, we are told to deploy using this command  remote_agent = client.agent_engines.create(     agent=app,     config={         "requirements": ["google-cloud-aiplatform [ agent_engines,adk ] "],         "staging_bucket": "STAGING_BUCKET",     } ) And that's was i bump into this error. To resolve it, I use the following code remote_agent = agent_engines.create(     agent,     requirements=[         "google-cloud-aiplatform [ agent_engines,langchain ] ==1.140.0",         "pydantic==2.12.3",         "cloudpickle==3.1.2"     ] ) And then i was ble to see that my agent being deployed. 

In this example, we will be using Azure verified module to create a storage account. So first we need  module cheapStorage 'br/public:avm/res/storage/storage-account:0.31.2' = {   name : 'storageDeployment'   params : {     // Required: Must be 3-24 lowercase letters/numbers and globally unique     name : 'stg ${ uniqueString ( resourceGroup (). id ) } '     location : resourceGroup (). location         skuName : 'Standard_LRS'     kind : 'StorageV2'     accessTier : 'Hot' // 'Hot' is better if you'll actually look at the files; 'Cool' is cheaper for long-term storage         allowBlobPublicAccess : false     networkAcls : {       defaultAction : 'Allow'       bypass : 'AzureServices'     }   } } // Output the primary endpoint so you can actually find your new toy output storageUri string = cheapStorage . outputs . primar...

In the previous post we creating azure storage account using bicep. Now we are going to create a service bus module and calling it.  So we will place our service bus under the path modules/servicebus/main.bicep and its contents are given here. This will create a service bus namespace and a queue @description('Name of the Service Bus namespace') param serviceBusNamespaceName string @description('Name of the Queue') param serviceBusQueueName string @description('Location for all resources.') param location string = resourceGroup().location resource serviceBusNamespace 'Microsoft.ServiceBus/namespaces@2022-01-01-preview' = {   name: serviceBusNamespaceName   location: location   sku: {     name: 'Standard'   }   properties: {} } resource serviceBusQueue 'Microsoft.ServiceBus/namespaces/queues@2022-01-01-preview' = {   parent: serviceBusNamespace   name: serviceBusQueueName   properties: {     lockDuration: 'PT5M'   ...

To create a storage account we can make our bicep code more re-usable by placing in into modules So we have this as our directory structure root (main.bicep) -> module -> storage -> (we call this main.bicep too)  root main.bicep contains the following code // main.bicep module storageModule './modules/storage/main.bicep' = {   name: 'storageDeploy-${uniqueString(resourceGroup().id)}'    params: {     storageSku: 'Standard_LRS'   } } // Accessing an output from the module output storageId string = storageModule.outputs.saName   And our storage module (modules/storage/main.bicep)  // main.bicep @description('Storage Account Name (must be globally unique)') param storageName string = 'store${uniqueString(resourceGroup().id)}' @description('The location for the resource') param location string = resourceGroup().location @description('Storage SKU') @allowed([   'Standard_LRS'   'Standard_GRS' ]) param storage...

This link provides lots of resource and sample code to work with Bicep  https://github.com/Azure/azure-quickstart-templates/tree/master/quickstarts

  Potential you can check its main branch for example for Qwen, it will be this here https://huggingface.co/Qwen/Qwen3.5-2B/tree/main And then you can see there's safetensor as shown below:-

 To create a cloud formation sqs, we can use the following code AWSTemplateFormatVersion : ' 2010-09-09 ' Description : ' Creates SQS Queue with specific User Access ' Parameters :   AccountId :     Type : String     Description : Your 12-digit AWS Account ID     Default : ' 00000000 '   UserName :     Type : String     Description : IAM User to grant access     Default : ' jeremydev ' Resources :   MySQSQueue :     Type : AWS::SQS::Queue     Properties :       QueueName : mytestsqs1       VisibilityTimeout : 300   MySQSQueuePolicy :     Type : AWS::SQS::QueuePolicy     Properties :       Queues :         - ! Ref MySQSQueue       PolicyDocument :         Version : ' 2012-10-17 '         Statement :           - Sid : Allo...

 First of all we need to run the following scripts to create our lambda function in python/  AWSTemplateFormatVersion : ' 2010-09-09 ' Description : ' AWS Lambda Function with IAM Role and Log Group ' Resources :   # 1. IAM Role for Lambda Execution   LambdaExecutionRole :     Type : AWS::IAM::Role     Properties :       AssumeRolePolicyDocument :         Version : ' 2012-10-17 '         Statement :           - Effect : Allow             Principal :               Service : [ lambda.amazonaws.com ]             Action : [ ' sts:AssumeRole ' ]       ManagedPolicyArns :         - arn:aws:iam::aws:policy/service-role/AWSLambdaBasicExecutionRole   # 2. CloudWatch Log Group (Prevents logs from being kept forever)   LambdaLogGroup :  ...

To create a EKS using cloud formation, we can use the following cloud formation code. This is not an EKS Auto AWSTemplateFormatVersion : ' 2010-09-09 ' Description : ' EKS Cluster and Managed Node Group ' Parameters :   ClusterName :     Type : String     Default : ' my-eks-cluster '   VpcId :     Type : AWS::EC2::VPC::Id   SubnetIds :     Type : List<AWS::EC2::Subnet::Id>     Description : ' Select at least two subnets in different AZs ' Resources :   # 1. IAM Role for the EKS Cluster (Control Plane)   EKSClusterRole :     Type : AWS::IAM::Role     Properties :       AssumeRolePolicyDocument :         Version : ' 2012-10-17 '         Statement :           - Effect : Allow             Principal :               Service : [ eks.amazonaws.com ] ...

 We can use the following command to view kubernetes tls secret. kubectl get secret/woolworths-wildcard-chain -n istio-ingress -o jsonpath="{.data['tls\.crt']}" | base64 --decode | openssl x509 -text -noout

In general we would have tar installed on a kubernetes pod. Sometimes when things are not happening, we can still copy to the container using the following command :-  Copy to container cat local-backup.tar.gz | kubectl exec -i -n cattle-resources-system \ rancher-backup-57c997f8cd-fxlb4 -c rancher-backup -- \ sh -c 'cat > /var/lib/backups/backup.tar.gz' Copying from container  kubectl exec -n cattle-resources-system rancher-backup-84576c498c-lpjh2 \ -c rancher-backup -- cat /var/lib/backups/default-location-recurring-backup-8723b2bd-e272-430e-89f6-14a133f8a417-2026-02-25T03-12-05Z.tar.gz \ > local-backup.tar.gz

When installing qwen code using the following command  curl -fsSL https://qwen-code-assets.oss-cn-hangzhou.aliyuncs.com/installation/install-qwen.sh | bash  it stuck in endless loop. This is due to installer not having permission to the /usr folder. To resolve it # 1. Create the directory for global packages mkdir -p ~/.npm-global # 2. Tell npm to use this new directory npm config set prefix '~/.npm-global' # 3. Open (or create) your shell profile file # If you use Bash (default on many Linux): echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc source ~/.bashrc Install gwen  npm install -g @qwen-code/qwen-code@latest And walla!