hamelsmu

Here is my personal opinion about the questions I posed in this tweet:

---

I think that fine-tuning is still very valuable in many situations. I’ve done some more digging and I find that people who say that fine-tuning isn't useful are indeed often working on products where fine-tuning isn't likely to be useful:

• They are making developer tools - foundation models have been trained extensively on coding tasks.
• They are building foundation models and testing for the most general cases. But the foundation models themselves are also being trained for the most general cases.
• They are building a personal assistant that isn’t scoped to any type of domain or use case and is essentially similar to the same folks building foundation models.

karpathy

"""
stable diffusion dreaming
creates hypnotic moving videos by smoothly walking randomly through the sample space

example way to run this script:

$ python stablediffusionwalk.py --prompt "blueberry spaghetti" --name blueberry

to stitch together the images, e.g.:
$ ffmpeg -r 10 -f image2 -s 512x512 -i blueberry/frame%06d.jpg -vcodec libx264 -crf 10 -pix_fmt yuv420p blueberry.mp4

amcchord

Lots of folks have asked me about my thoughts on Kaseya’s purchase of Datto:

Until recently I have been very much telling everyone to assume positive intent. Kaseya is looking to continue Datto’s success. In general, people don’t spend 6+ Billion dollars on something they intend to break. Change is inevitable and mistakes will be made but by and large Datto should expect to continue to be the company that is loved by employees and customers alike.

This past week, many current members of the Datto team have reached out deeply dismayed. There is a concern that the current trajectory from Datto’s new owners will snuff the flame that makes Datto a place to come “Do your life’s work.”

I am not associated with the company any more so my understanding of the specifics comes entirely second hand. Current team members have reached out and described the following:

- Sidelining Employee Resource Groups that support under represented people at the company

rochacbruno

from typing import Union, List
from dataclasses import dataclass


class Validations:
    def __post_init__(self):
        """Run validation methods if declared.

        The validation method can be a simple check
        that raises ValueError or a transformation to

osanseviero

/*
 * Converts table from [[Header0, Header1, Header2], [Column0Val0, Column1Val0, Column2Val0], ...]
 * to {Header0: [ColumnVal0, ...], Header1: [Column1Val0, ...], Header2: [Column2Val0, ...]}
 */
function convertTableToData(table) {
    transposed = table[0].map((_, colIndex) => table.map(row => row[colIndex]));
    result = {}
    for(var i = 0; i < transposed.length; i++) {
      header = transposed[i][0]
      result[header] = transposed[i].slice(1, 4).map(String);

rwightman

""" A simple U-Net w/ timm backbone encoder

Based off an old version of Unet in https://github.com/qubvel/segmentation_models.pytorch

Hacked together by Ross Wightman
"""

from typing import Optional, List

import torch

pzelasko

import sys
import torch
from torch.utils.data import dataloader
from torch.multiprocessing import reductions
from multiprocessing.reduction import ForkingPickler

default_collate_func = dataloader.default_collate


def default_collate_override(batch):

ramdesh

import s3fs
import zipfile
import tempfile
import numpy as np
from tensorflow import keras
from pathlib import Path
import logging


AWS_ACCESS_KEY="aws_access_key"

egocarib

# This code adapted from https://github.com/python-pillow/Pillow/issues/4644 to resolve an issue
# described in https://github.com/python-pillow/Pillow/issues/4640
#
# There is a known issue with the Pillow library that messes up GIF transparency by replacing the
# transparent pixels with black pixels (among other issues) when the GIF is saved using PIL.Image.save().
# This code works around the issue and allows us to properly generate transparent GIFs.

from typing import Tuple, List, Union
from collections import defaultdict
from random import randrange

scottire

import numpy as np
import panel as pn
import holoviews as hv 
hv.extension("bokeh")
# hv.extension('matplotlib')
from holoviews.streams import Stream, Params
from scipy.io import wavfile
from scipy.signal import spectrogram

rate, data = wavfile.read('/filepath.wav')