chu-tianxiang

import time
import torch
import torch.utils.benchmark as benchmark
import triton
import triton.language as tl


@triton.jit
def _fwd_kernel(
    Q1, Q2, K1, K2, V, sm_scale,

chu-tianxiang

# Adapted from the triton implementation of flash-attention v2
# https://github.com/openai/triton/blob/main/python/tutorials/06-fused-attention.py
import time
import torch
import torch.utils.benchmark as benchmark
import triton
import triton.language as tl


@triton.jit

chu-tianxiang

import numpy as np
import torch
import jax
from tqdm import tqdm

from model import LanguageModelConfig, TransformerConfig, QuantizedWeight8bit as QW8Bit
from runners import InferenceRunner, ModelRunner, sample_from_model

CKPT_PATH = "./checkpoints"