All posts

Attention dilution (also called context dilution) is one of the fundamental limitations of transformer-based LLMs when dealing with long contexts or extended agent memory.

How many of these terms do you actually recognize?

From input to output, a prompt generally goes through seven steps: request packaging, tokenization, inference scheduling, prefill, and decode before the result is returned.

ChatGPT Stats ChatGPT Growth ChatGPT Revenue

Over the next 12 to 24 months, the differentiator among engineers will shift from mastery of programming languages like Rust, Go, or Python, or the volume of code produced, to the...

Hyperparameters are external settings chosen before training, such as the learning rate or regularization strength.

As large language models (LLMs) scale up, researchers have begun to notice a growing imbalance between model size and the availability of high-quality training tokens. The...

In large-language-model (LLM) inference serving contexts, once the model compute becomes sufficiently fast, the performance bottleneck often shifts to the key-value (KV) cache...

Reflection is related to agent self-improvement or reasoning feedback loops.

[x] Independent deployable services - Each agent can scale horizontally (e.g., analysisservice replicas) - You can version and deploy agents independently

Its advantages over traditional sequential chains are evident in two areas:

1. Objective 2. Environment Setup

MCP Server Hub Currently, our different projects are using various MCP servers. To streamline and unify the process, we plan to implement a HUB MCP server that can handle multiple...

Tools in Large Language Models (LLMs) Tools enable large language models (LLMs) to interact with external systems, APIs, or data sources, extending their capabilities beyond text...

LangChain Invoke Retry Logic LLM call is not stable and may fail due to network issues or other reasons, therefore, retry logic is necessary.

| Feature | stdio | sse (Server-Sent Events) | streamable-http | |--------------------------|------------------------------------------|--------------------------------------------...

Out: None [Step 1: Duration 146.87 seconds| Input tokens: 2,113 | Output tokens: 923] ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Step 2 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ─ Executing...

Step-by-Step Guide: Building an MCP Server using Python-SDK, AlphaVantage & Claude AI Model Context Protocol (MCP) lab

Retrieval-Augmented Generation (RAG) is a powerful approach that combines retrieval and generation to produce high-quality responses. However, the quality of the final response can...

You start by creating a Modelfile, which acts as a key to unlock any GGUF model you want to use.

Learning never exhausts the mind         ― Leonardo da Vinci

Skyvern ScrapegraphAI Crawl4AI Reader Firecrawl Markdowner

|Feature| LangGraph| AutoGen| |---|---|---| |Core Concept| Graph-based workflow for LLM chaining| Multi-agent system with customizable agents| |Architecture| Node-based computation...

AutoGen is a framework for creating multi-agent AI applications that can act autonomously or work alongside humans.

If you find this in your VSCode, congratulations! You have successfully set up Ollama for code generation and assistance in Visual Studio Code. alt text

%%{init: { 'look':'handDrawn' } }%%

```python linenums="1" spark = ( SparkSession.builder.master("local[]").appName("test").getOrCreate() ) d = [ Event(1, "abc"), Event(2, "ddd"), ]

My previous spark project is scala based and I use IDEA to compile and test conveniently.:smile::smile::smile: Databricks Job nice UI save your time to create JAR job.

:bulb: It will extend your function behaviors during runtime.

This video is helpful to understand it. type:video

Reflex (pynecone) Reflex is a library to build full-stack web apps in pure Python. Repo Video type:video

I have enrolled in a private Snowflake Data Science Training. Let me list what I learned from it.

We can use internal runpy to execute different moduls in our project.

```python linenums="1" title="myclient.py"

Problem: How to introduce ml-based production/features to cross-functional teams.

bin/spark-submit \ master k8s://https://192.168.99.100:8443 \ deploy-mode cluster \ name spark-pi \ class org.apache.spark.examples.SparkPi \ conf spark.driver.cores=1 \ conf...

Recently I'm working in Azure to implement ETL jobs. The main tool is ADF (Azure Data Factory). This post show some solutions to resolve issue in my work.

scala ref create dataframe

```txt master MASTERURL --> 运行模式 例：spark://host:port, mesos://host:port, yarn, or local.

PROCESSLOCAL data is in the same JVM as the running code. This is the best locality possible NODELOCAL data is on the same node. Examples might be in HDFS on the same node, or in...

import airflow from airflow.models import DAG from airflow.operators.pythonoperator import PythonOperator

Whitening Transformation

Recently reading a blog Structured Streaming in PySpark It's implemented in Databricks platform. Then I try to implement in my local Spark. Some tricky issue happened during my...

Batch Normalization is one of important parts in our NN.

Vanilla gradient descent, aka batch gradient descent, computes the gradient of the cost function w.r.t. to the parameters θ

Repos Repo List language link