andreaskoepf

local function pca(X)
  -- PCA -------------------------------------------------------------------------
  -- X is m x n
  local mean = torch.mean(X, 1) -- 1 x n
  local m = X:size(1)
  local Xm = X - torch.ones(m, 1) * mean
  Xm:div(math.sqrt(m - 1))
  local v,s,_ = torch.svd(Xm:t())
  s:cmul(s) -- n
  

andreaskoepf

function YUV422toYUV(input, w, h)
  local yuv_result = torch.ByteTensor(3, h, w)
  local x = input:view(h, w, 2)
  yuv_result[1] = x[{{},{},1}] -- copy Y part
  local u = yuv_result[2]
  local v = yuv_result[3]
  local uv = x[{{},{},2}]:reshape(h,w/2,2)
  local u_ = uv[{{},{},1}]
  local v_ = uv[{{},{},2}]
  u:view(h,w/2, 2)[{{},{},1}] = u_

andreaskoepf

--[[
torch.chunk2()
returns a table with nChunk entries, even in the case that the tensor has < nChunk entries in the specified dimension. 

Behaviour of the originial torch.chunk() function:

th> torch.rand(11):chunk(5)
{
  1 : DoubleTensor - size: 3
  2 : DoubleTensor - size: 3

andreaskoepf

using Microsoft.Extensions.DependencyModel;
using System;
using System.Linq;
using System.Runtime.Loader;
using System.Reflection;
using System.Collections.Generic;

namespace ConsoleApp1
{
    class Program

andreaskoepf

# adapted from: https://github.com/lucidrains/vit-pytorch/blob/main/vit_pytorch/vit.py

import torch
import torch.nn.functional as F
from torch import nn
from torch.utils import checkpoint

from einops import rearrange, repeat

andreaskoepf

# adapted from: https://github.com/lucidrains/vit-pytorch/blob/main/vit_pytorch/vit.py

import torch
import torch.nn.functional as F
from torch import nn
from torch.utils import checkpoint

from einops import rearrange, repeat

import triton

andreaskoepf

# sent to me by tju01, thx

# install base tools
apt update
apt install protobuf-compiler libssl-dev gcc pkg-config g++ make 

# install rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source "$HOME/.cargo/env"

andreaskoepf

from typing import Optional
import torch


def precompute_freqs_cis(
    dim: int, end: int, theta: float = 10000.0, scaling_factor: float = 1.0
) -> torch.Tensor:
    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2).float() / dim))
    t = torch.arange(end, device=freqs.device).float() / scaling_factor  # type: ignore
    freqs = torch.outer(t, freqs).float()  # type: ignore

andreaskoepf

import argparse
import sys

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(
        description="Push checkpoints in HF transformers format to the Huggingface Hub.",

andreaskoepf

You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct.
If you don't know the answer to a question, please don't share false information."