corbt

Strategy                       | Relative Throughput  | Time (s)     | Cost ($/M tokens) 
----------------------------------------------------------------------------------------
Unsloth                        | 2.17                 | 3.83         | $0.0188           
Unsloth+PEFT                   | 1.58                 | 5.27         | $0.0259           
Transformers+Liger             | 1.14                 | 7.28         | $0.0358           
vLLM                           | 1.00                 | 8.31         | $0.0409           
Transformers                   | 0.97                 | 8.54         | $0.0420           
Transformers+Liger+PEFT        | 0.84                 | 9.85         | $0.0484           
Transformers+PEFT              | 0.74                 | 11.26        | $0.0554           

hamelsmu

Question: Should I avoid using RAG for my AI application after reading that "RAG is dead" for coding agents?

Many developers are confused about when and how to use RAG after reading articles claiming "RAG is dead." Understanding what RAG actually means versus the narrow marketing definitions will help you make better architectural decisions for your AI applications.

Answer: The viral article claiming RAG is dead specifically argues against using naive vector database retrieval for autonomous coding agents, not RAG as a whole. This is a crucial distinction that many developers miss due to misleading marketing.

RAG simply means Retrieval-Augmented Generation - using retrieval to provide relevant context that improves your model's output. The core principle remains essential: your LLM needs the right context to generate accurate answers. The question isn't whether to use retrieval, but how to retrieve effectively.

For coding

stevebauman

import { defineConfig, loadEnv } from 'vite';

// ...

export default defineConfig(({ mode }) => {
    const env = loadEnv(mode, process.cwd());

    const { protocol, hostname } = new URL(env.VITE_URL);

    const root = hostname.split('.').slice(-2).join('\\.');

perfectbase

/* eslint-disable @typescript-eslint/no-explicit-any */
import { type TRPCQueryOptions } from '@trpc/tanstack-react-query';
import { unstable_noStore } from 'next/cache';
import { Fragment, Suspense, type ReactNode } from 'react';
import { ErrorBoundary } from 'react-error-boundary';
import { HydrateClient, prefetch as prefetchTRPC } from '@/trpc/server';

type AwaitProps<T> =
  | {
      promise: Promise<T>;

joseph-crowley

Below is a compressed yet complete reference for quickly integrating each shadcn component. Assumption: you already have the files from your question in @/components/ui/*.tsx and can import them directly. All components accept typical React props plus any Radix/3rd-party props. Adjust styling and props as needed.Do not rewrite any of the code for the shadcn components.

---

1. Accordion

Import

import {
  Accordion,
  AccordionItem,

kalomaze

# the "verifiers" repository is a clean implementation of templated GRPO reinforcement learning training environments
# this is a generic set of "install from scratch" commands complete with a deepspeed z3 config that i have been using when i spin up nodes
# it will run on the gsm8k example w/ default batch size & generation size (8), and the 8th GPU is used for vllm generations
# qwen 14b full finetuning will run on this configuration too without LoRA or CUDA OOM, at least for the gsm8k task's context sizes + generation lengths
# hyperparameters are controlled by `verifiers/utils/config_utils.py`; i have been preferring extreme grad clipping (between 0.001 and 0.01) and low beta (under 0.01)

# NOTE FEB 27: examples have moved into `verifiers/examples` not `/examples`

cd /root
mkdir boom

sayakpaul

from google import genai
from google.genai import types
import typing_extensions as typing
from PIL import Image
import requests
import io
import json
import os

qunash

Chillee

import torch
from torch.nn.attention.flex_attention import create_block_mask, flex_attention
torch.set_default_device('cuda')

q, k, v = [torch.randn(8, 8, 1024, 64, requires_grad=True) for _ in range(3)]

causal_mask = create_block_mask(lambda b, h, q_idx, kv_idx: q_idx >= kv_idx, None, None, 1024, 1024)
uncausal_mask = create_block_mask(lambda b, h, q_idx, kv_idx: q_idx < kv_idx, None, None, 1024, 1024)
ref_out = flex_attention(q, k, v)
causal_out, causal_lse = flex_attention(q, k, v, block_mask=causal_mask, return_lse=True)

willccbb

# train_grpo.py
#
# See https://github.com/willccbb/verifiers for ongoing developments
#
"""
citation:

@misc{brown2025grpodemo,
  title={Granular Format Rewards for Eliciting Mathematical Reasoning Capabilities in Small Language Models},
  author={Brown, William},