Google Gemini

  • Gemini is a family of generative AI models developed by Google DeepMind that is designed for multimodal use cases.
  • The Gemini API gives you access to the Gemini Pro and Gemini Flash models.
  • Calling functions from Gemini
    • The Vertex AI Gemini API provides a unified interface for interacting with Gemini models.
    • Function calling lets developers create a description of a function in their code, then pass that description to a language model in a request. 
    • The response from the model includes the name of a function that matches the description and the arguments to call it with.
    • Function calling is similar to Vertex AI Extensions . The difference between them is that function calling returns JSON data with the name of a function and the arguments to use in your code, whereas Vertex AI Extensions returns the function and calls it for you.
    • Define a search_db function with parameters for search terms.
    • Create a get_weather function that takes a location as input.
    • Function calling bridges the gap between human language and the structured data needed to interact with external systems.
    • Use the Vertex AI Gemini API to interact with the Gemini Flash.
  • Gemini models are multimodal they can understand text, audio, and video.
    • prompts can contain a combination of text, images, and audio.
  • To count number of tokens in your prompt use count token API.
    • Input can be up to 104 8576 tokens.
    • 60 minutes of video without audio.
    • 45 minutes of video with audio.
    • 84 hours of audio.
    • 3000 documents.
    • 3000 images.
  • Features of Gemini Models.
    • Grounding with google search to base responses on real time data.
    • Structured output in specific schema.
    • Function calling- provide functions it can call.
    • Code Execution - generate and execute code.
    • Thinking training to generate thinking process.
  • Gemini Pro
    • Complex, problem-solving, and logical data.
    • Nuanced understanding across diverse inputs.
    • Simultaneous multimodal processing and reasoning
  • Gemini Flash
    • Optimised for speed and cast.
    • Low latency and high throughput responses.
    • Designed for cost sensitive, high volume AI applications.
  • Gemini Flash Lite
    • Stripped down, optimized version of Flash.
    • Designed for rapid, inexpensive text generation at scale.
  • Limitations of Gemini
    • Spatial reasoning
    • Counting
    • Follow Complex Instructions
    • Hallucinations
    • Medical Uses
  • Building Generative AI solutions using Model Garden Models and API's

Comments

Post a Comment

Popular posts from this blog

High Level Diagrams(HLD's)

Image
High Level Diagram for Building a Serverless application using Step Functions, API Gateway, Lambda, and S3 in AWS HLD for Building and Troubleshooting a Serverless Web
Read more

AWS Summaries

Database Summary  RDS(OLTP) SQL My SQL Post Gre SQL Oracle Aurora Maria DB Dynamo DB(No SQL) Redshift OLAP Elastic cache Memcached Redis RDS runs on virtual machines We cannot log into this operating systems however Patching of our RDS Operating system and DB is Amazons responsibility. RDS is not server-less. Aurora surverless is seven less which is an exception. Dynamo DB is server-less. There are two types of backups for RDS Automated backups Database snapshots Read replicas Can be multi availability zones. Use to increase performance Must have backups turned on. Can be in different regions. Can be MySQL,PostgreSQL,Maria DB, Oracle, Aurora, SQL Server is not supported. Can be promoted to master, this will break read replica. Multi availability zone Used for DRG only and not performance. You can force a fail over from one availability zone to another by re-booting the RDS instance. Encryption at Rest Is supported for My SQL, Oracle, SQL server, Postgre SQL, Maria DB and Aurora. En...
Read more

Exam Tips

AWS CLI Least Privilege Always give your users, the minimum amount of access required to do the job. Use Groups Create IAM policy groups and assign your users to the groups. Group permissions are assigned using IAM policy documents. Secret access key You only see this ones. If you lose it you can delete the access key and secret access key and regenerate them. You need to run AWS configure again. Don’t share key pairs Each developer should have their own access Key ID and secret access Key. Just like passwords, they should not be shared. Supports Linux, Windows, macOS You can install CLI on your Mac, Linux, or Windows PC. You can also use it on EC2 instances. AWS CLI Pagination If you see errors like “ timed out”, or errors related to too many results being returned. Adjust the pagination of CLI results to avoid errors generated by too many results. aws S3 api list-objects - - bucket my-bucket  - - page-size 100. The CLI still retrieves the full list, but performs a large number of...
Read more