pieFi

PowerInfer‑2 achieves up to 29× speedups over llama.cpp and 13× over LLMFlash by leveraging neuron‑level pipelines and NPU‑centric prefill optimization.

Performance Evaluation of PowerInfer‑2: Offloading, Prefill, and In‑Memory Efficiency

This article outlines key PELT methods (BitFit, Adapter, Prefix-Tuning) and motivates UNIPELT, a gated hybrid combining them for robust LM tuning. 

The Tuning Trifecta: UNIPELT’s Gated Symphony of BitFit, Adapter, and Prefix-Tuning

This section highlights vAttention's ability to add dynamic memory allocation support to unmodified FlashAttention and FlashInfer prefill kernels

vAttention Performance & Portability for LLM Prefill Phase

This section details the evaluation methodology for vAttention, assessing its performance, portability, and memory efficiency in both prefill and decode phases

Evaluation of vAttention for LLM Inference: Prefill and Decode Performance

We explore the effectiveness of vLLM for the case a prefix is shared among different input prompts

How Effective is vLLM When a Prefix Is Thrown Into the Mix?

Avoid using prefixes on collections whenever possible.

Code Smell 257 - Simplifying Property Names By Removing Prefixes

Discover how Dramatron's creativity is fueled by diverse prompt sets, including Medea's tragedy from Ancient Greece and Sci-Fi inspirations like Star Wars.

Diverse AI Prompt Prefixes for Scriptwriting: From Star Wars to Ancient Drama

Learn how to add a prefix or suffix to each new line in JavaScript with our step-by-step tutorial. Create a user-friendly online tool using JS.

How to Add Prefix or Suffix to Each New Line in JavaScript

A trie (pronounced as “try”) or prefix tree is a tree data structure used to efficiently store and retrieve keys in a dataset of strings.

How to Implement Trie (Prefix Tree) - Blind 75 LeetCode Questions

In this post, we will look at prefix sums and how they can be used to solve a common coding problem, that is, calculating the sum of an array (segment). This article will use Java for the code samples but the concept should apply to most programming languages.

Prefix Sums and How They Can be Used to Solve Coding Problems

GraphQL’s concept spans from data store to client. What happens along the way? Tooling in the <a href="https://hackernoon.com/tagged/graphql" target="_blank">GraphQL</a> world has come a long way and spinning up an application is easier than ever. By stepping through each component, you’ll have a better understanding of GraphQL as a whole.

Piecing Together GraphQL

As much as I want to start using <a href="https://hackernoon.com/tagged/javascript" target="_blank">JavaScript</a> right away, and create applications, I know that I won’t be able to fully grasp the language unless I understand the fundamentals. Consequently, I’ve been following the chapters from <a href="https://javascript.info" target="_blank">https://javascript.info</a> which has been a great source so far.

JavaScript — Back to Basics: Prefix vs. Postfix

Raw measurements don’t tell the full story. 

Why Measuring Time is Not Enough: a Practical Roofline Model for ML Training

How ShareChat scaled its ML feature store 1000× using ScyllaDB, smarter data modeling, and caching—without scaling the database.

How ShareChat Scaled their ML Feature Store 1000X without Scaling the Database

HackerNoon curates 179 standout headlines across AI, Web3, coding, startups, and more—your fastest path to high-value tech reads.

179 Super Compelling HackerNoon Headlines

Explore the differences between HTMX and Lightview hypermedia. Learn how to choose between pure HDA architecture and Lightview’s multiple paradigms. 

Building with Hypermedia: HTMX's Purity and Lightview's Flexibility.

From mainframe DFSMShsm to cloud storage classes: a practical history of HSM, ILM, tiering, recall, and the products that shaped modern archives.

HSM: The Original Tiering Engine Behind Mainframes, Cloud, and S3

OPT provides open access to GPT-3 scale language models, enabling responsible AI research with lower compute costs and reproducible experiments.

OPT-175B is Comparable to GPT-3 While Requiring Only 1/7th the Carbon Footprint 

Claude Code Launches Teleport Workflow: Start Anywhere, Continue Everywhere

Translate a TypeScript compiler AST (via protobuf/gRPC) into your own generalized model.

The Elephant in the Room: Why This AST Translation Must Be Code-Generated

This article will discuss compression in the Big Data context, covering the types and methods of compression

Compression in Big Data: Types and Techniques

IPv6 breaks digital ad measurement. Learn how IPinfo’s research-driven, active-measurement model restores accuracy across CTV and all channels.

IPv6 and CTV: The Measurement Challenge From the Fastest-Growing Ad Channel

CodeGrok MCP replaces grep with semantic code search for AI agents. AST parsing + vector embeddings = 10x better context efficiency. 100% local

CodeGrok MCP: Semantic Code Search That Saves AI Agents 10x in Context Usage

Ollama is an open-source platform for running and managing large-language-model (LLM) packages entirely on your local machine.

PIEFI

piefi.io

10-50 emps

Since 2021

pieFi's stories on HackerNoon

Performance Evaluation of PowerInfer‑2: Offloading, Prefill, and In‑Memory Efficiency

The Tuning Trifecta: UNIPELT’s Gated Symphony of BitFit, Adapter, and Prefix-Tuning

vAttention Performance & Portability for LLM Prefill Phase

Evaluation of vAttention for LLM Inference: Prefill and Decode Performance

How Effective is vLLM When a Prefix Is Thrown Into the Mix?

Code Smell 257 - Simplifying Property Names By Removing Prefixes

Diverse AI Prompt Prefixes for Scriptwriting: From Star Wars to Ancient Drama

How to Add Prefix or Suffix to Each New Line in JavaScript

How to Implement Trie (Prefix Tree) - Blind 75 LeetCode Questions

Prefix Sums and How They Can be Used to Solve Coding Problems

Piecing Together GraphQL

JavaScript — Back to Basics: Prefix vs. Postfix

Why Measuring Time is Not Enough: a Practical Roofline Model for ML Training

How ShareChat Scaled their ML Feature Store 1000X without Scaling the Database

179 Super Compelling HackerNoon Headlines

Building with Hypermedia: HTMX's Purity and Lightview's Flexibility.

HSM: The Original Tiering Engine Behind Mainframes, Cloud, and S3

OPT-175B is Comparable to GPT-3 While Requiring Only 1/7th the Carbon Footprint

Claude Code Launches Teleport Workflow: Start Anywhere, Continue Everywhere

The Elephant in the Room: Why This AST Translation Must Be Code-Generated

Compression in Big Data: Types and Techniques

IPv6 and CTV: The Measurement Challenge From the Fastest-Growing Ad Channel

CodeGrok MCP: Semantic Code Search That Saves AI Agents 10x in Context Usage

Complete Ollama Tutorial (2026) – LLMs via CLI, Cloud & Python

Representing this company? Click here to claim and customize this page!

Reviews About pieFi

pieFi WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps

Performance Evaluation of PowerInfer‑2: Offloading, Prefill, and In‑Memory Efficiency