willccbb

# train_grpo.py
#
# See https://github.com/willccbb/verifiers for ongoing developments
#
import re
import torch
from datasets import load_dataset, Dataset
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import LoraConfig
from trl import GRPOConfig, GRPOTrainer

cyphunk

// Fork & examples for the one-line version by @vladimir-ivanov:
//let softmax = (arr) => (index) => Math.exp(arr[index]) / arr.map(y => Math.exp(y)).reduce((a, b) => a + b);
//
// Also see comments for improvements

function softmax(arr) {
    return arr.map(function(value,index) { 
      return Math.exp(value) / arr.map( function(y /*value*/){ return Math.exp(y) } ).reduce( function(a,b){ return a+b })
    })
}