abodacs

abodacs

# train_grpo.py
import re
from datasets import load_dataset, Dataset
from transformers import AutoTokenizer
from peft import LoraConfig
from trl import GRPOConfig, GRPOTrainer

# Load and prep dataset

SYSTEM_PROMPT = """

abodacs

Classify user search queries as either "Good Google Search Query" or "Bad Google Search Query" based on their likelihood of yielding relevant and helpful results from Google Search.

Input: User search query (text string).

Output: Classification label:
    * Good Google Search Query: The query is likely to be effectively answered by Google Search.
    * Bad Google Search Query: The query is unlikely to be effectively answered by Google Search.  Further categorize "Bad" queries into subtypes for better understanding and classifier training (optional but highly recommended):
        * Chit-Chat/Conversational/Social
        * Personal/Subjective/Opinion-Based (Un-searchable)
        * Vague/Ambiguous/Lacking Specificity

abodacs

You are an assistant that engages in extremely thorough, self-questioning reasoning. Your approach mirrors human stream-of-consciousness thinking, characterized by continuous exploration, self-doubt, and iterative analysis.

## Core Principles

1. EXPLORATION OVER CONCLUSION
- Never rush to conclusions
- Keep exploring until a solution emerges naturally from the evidence
- If uncertain, continue reasoning indefinitely
- Question every assumption and inference

abodacs

import multiprocessing

manager = multiprocessing.Manager()
all_hashes_set = manager.dict()
def deduplicate(examples, all_hashes_set):
    print(len(all_hashes_set))
    input_ids = examples['input_ids']
    hashes = [
        hash(tuple(input_ids[i]))
        for i in range(len(input_ids))

abodacs

# Train GPT-2 in five minutes -- for free
#
# ```bash
# pip install modal
# modal setup
# modal run wrapper.py
# ```
#
# Note that the end-to-end latency the first time is more like 25 minutes:
# - five minutes to install Torch (rip)

abodacs

import torch
from diffusers import FluxPipeline
from torch import nn


class ModelOffloaderV2:
    def __init__(self, model: nn.Module, record_stream: bool = False):
        # move model to pinned memory. keep a model copy in CPU pinned memory.
        for p in model.parameters():
            p.data = p.data.cpu().pin_memory()

abodacs

require("dotenv").config();
const H = require("highland");
const axios = require("axios");
const fs = require("fs").promises;

const exportDirectory = "./export";
const apiUrl = "https://api.intercom.io";

// config axios for the authorized API request
const apiClient = axios.create({

abodacs

var $debugHelper = $debugHelper || {};
$debugHelper = function () {
    var href = "lib/debugger.css";
    var addCss = function () {
        if (styleStyleIsLoaded(href) === true) {
            return;
        }
        const head = document.head;
        const link = document.createElement("link");
        link.type = "text/css";

abodacs

# REQUIRES torchao, torch nightly (or torch 2.5) and transformers

from transformers import AutoTokenizer, AutoModelForCausalLM, TorchAoConfig
from transformers import TextStreamer
import torch
from tqdm import tqdm
import os
os.environ["TOKENIZERS_PARALLELISM"] = "false"  # To prevent long warnings :)

torch.set_float32_matmul_precision('high')