DSPy is a declarative Python framework for building modular, maintainable, and portable AI systems. Unlike traditional prompt engineering, DSPy lets you compose AI software from natural-language modules and structured code, enabling fast iteration and robust integration with a wide range of language models (LMs) and learning strategies.
| # 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 |
| 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) |
| import argparse | |
| import random | |
| import sys | |
| from transformers import AutoModelForCausalLM, AutoTokenizer, DynamicCache | |
| import torch | |
| parser = argparse.ArgumentParser() | |
| parser.add_argument("question", type=str) | |
| parser.add_argument( |
| import numpy as np | |
| import pandas as pd | |
| import statsmodels.api as sm | |
| import statsmodels.formula.api as smf | |
| import matplotlib.pyplot as plt | |
| # Function to simulate returns and signal | |
| def simulate_returns_and_signal(m_symbols, n_samples, seed): | |
| # Parameters | |
| r = 0 # Average return |
| import argparse | |
| import dataclasses | |
| import functools | |
| import json | |
| import logging | |
| import math | |
| import os | |
| import re | |
| import time |
| from typing import Tuple | |
| import jax | |
| import jax.numpy as jnp | |
| import chex | |
| from model import xmr_early_exit | |
| from draft_model import xmr_early_exit as sml_xmr_early_exit | |
| # feed-forward function for reward and value prediction networks | |
| def mlp(x: jax.Array, w1: jax.Array, w2: jax.Array, w3: jax.Array) -> jax.Array: |
| 08/01/2024 15:26:03 - WARNING - root - $$$$ Type: <class 'list'> <class 'langchain_core.documents.base.Document'> Total Documents: 119 | |
| 08/01/2024 15:26:10 - WARNING - root - ############# Starting the base retrieval ############################### | |
| 08/01/2024 15:26:11 - INFO - root - Document 1: | |
| Reasoning CLEVR, NLVR, VisualMRC | |
| Translation Multi30k | |
| VQA ActivityNet-QA, DocVQA, GQA, iVQA, | |
| MSRVTT-QA, MSVD-QA, OCR-VQA, ST- | |
| VQA, ViQuAE, VQAv2, Visual Dialog | |
| Table 10. The SFT blend we used during the ablation study. |
| import os | |
| import requests | |
| import numpy as np | |
| import logging | |
| from datetime import datetime, timedelta | |
| import json | |
| import logging | |
| import asyncio | |
| import time | |
| import concurrent.futures |
| from langchain_community.document_loaders import PyPDFLoader | |
| from langchain_text_splitters import RecursiveCharacterTextSplitter | |
| from langchain_community.vectorstores import FAISS | |
| import os | |
| import requests | |
| import numpy as np | |
| import faiss | |
| from mistralai.client import MistralClient | |
| from mistralai.models.chat_completion import ChatMessage |