Rafa Sánchez - The Stochastic Parrot Blog

Collection of Arxiv papers of the Gemma model family

A collection of Arxiv papers describing main innovations of Gemma model family.

Remote to Vertex AI Workbench Instances over an IAP tunnel with VS Code

VS Code can be used as frontend with Vertex AI Workbench Instances as remote, using an SSH connection and IAP without exposing an external IP address.

Vertex AI Workbench Comparison

Vertex AI Workbench and comparison with similar solutions, like Colab Enterprise

Generative AI - Deploy and inference of Llama 2 on Vertex AI Prediction

Deploy Llama 2 in Vertex AI Prediction with ctransformers

Generative AI - Q&A with semantic answering on large scanned documents

Gradio app containing Question and Answer with semantic answering on large scanned documents with Vertex AI LLM, Chroma, LangChain and Document AI OCR in batch mode

Generative AI - PaLM-2 model deployment with Cloud Run

Learn how to deploy a Gradio app in Cloud Run that calls a PaLM-2 model

Vector databases

Comparison between vector databases like Pinecone, Chroma, open-source ScaNN, Weviate and Vertex AI Matching Engine. Applications with LLMs.

Google high-level FAQ around Generative AI

Frequent questions about Generative AI in Google

GPT family review and Azure endpoints

Announcement and details around Generative Pretrained Transformer (GPT) model family

A review of inference servers for LLMs

A review of inference servers that can be used to deploy Large Language Models

Fine-tuning FLAN-T5 XXL with DeepSpeed and Vertex AI

Learn how to fine-tune a FLAN-T5 XXL model in Vertex AI, using the DeepSpeed library. The model will be fine-tuned with 8xA100 GPUs, and stored in Vertex AI Model Registry

Deploy FLAN-T5 XXL in Vertex AI Prediction

Learn how to deploy a FLAN-T5 XXL model in Vertex AI. The model will be downloaded and embedded in the custom prediction image, and deployed in an endpoint with 1xA100 GPU.

Fine-tuning Flan-T5 Base and online deployment in Vertex AI

Learn how to fine-tune and deploy a Flan-T5 Base model using the SAMSum dataset (summary of conversations in English) in Vertex AI

About stochasticity in LLMs

Autoregression, Generation and Autoencoding differences, as well as determinism vs stochasticity in LLMs

Definition of LLM

What is exactly an LLM ? Which models fit into this term ? Is a generative AI model an LLM ?

The JAX framework

JAX is not only a GPU-backed NumPy. Also provides Autograd and XLA for high-performance machine learning research

Document AI Technical FAQ

Most common technical questions around Document AI service in Google Cloud

Google OCR and Document AI

A review of OCR alternatives available in Google and differences between Vision API OCR and Document AI OCR

Open source tools for document preprocessing

Python open source tools to preprocess documents before calling Document AI services

How to install VerneMQ MQTT broker in GKE

How to install VerneMQ MQTT broker in GKE

Setting machine type in a Vertex AI Pipelines component

Unlike GPU/RAM, setting the machine type in a pipeline component is not so simple as in a Vertex AI Custom job. You can use create_custom_training_job_from_component to set…

On AutoML Translate and BLEU

When evaluating a new test set with AutoML Translate, it can only be done through the UI, and not with the API. Recommendation is to use always NLTK to evaluate

Using keras.mized_precission

Computer Vision

Using keras.mixed_precission in image segmentation

Formularios con Google Document AI

Structured Data

Análisis de documentos PDF y extracción de campos de formularios (Form parser) con Google Document AI

AutoML NLP Classifier with a large confussion matrix

Google Cloud AutoML NLP Classifier with a confussion matrix larger than 10x10

Transfer data in/out of Colab

Transfer data from from Google Drive, GCS and other external sources in/out of Colab