geniuspatrick

# Copyright 2023 UC Berkeley Team and The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,

geniuspatrick

Make 🤗 D🧨ffusers run on MindSpore

State-of-the-art diffusion models for image and audio generation in MindSpore. We've tried to provide a completely consistent interface and usage with the huggingface/diffusers. Only necessary changes are made to the huggingface/diffusers to make it seamless for users from torch.

Noise Scheduler

🤗 Diffusers provides many scheduler functions for the diffusion process. A scheduler takes a model's output (the sample which the diffusion process is iterating on) and a timestep to return a denoised sample. The timestep is important because it dictates where in the diffusion process the step is; data is generated by iterating forward n timesteps and inference occurs by propagating backward through the timesteps. Based on the timestep, a scheduler may be discrete in which case the timestep is an int or continuous in which case the timestep is a float.

geniuspatrick

#!/bin/bash

export RANK_SIZE=4

echo "Command: $@"

# trap SIGINT to execute kill 0, which will kill all processes
trap 'kill 0' SIGINT
for ((i = 0; i < ${RANK_SIZE}; i++)); do
    export RANK_ID=$i

geniuspatrick

# Basic
set -g default-terminal "screen-256color"  # enable 256 color
set -g mouse on  # enable mouse
set -g history-limit 50000

# Use Alt-arrow keys to switch panes
bind -n M-Left select-pane -L
bind -n M-Right select-pane -R
bind -n M-Up select-pane -U
bind -n M-Down select-pane -D

geniuspatrick

新服务器配置记录

零：安装Shell

a. Set Mirrors

首先备份配置文件：

sudo cp -a /etc/apt/sources.list /etc/apt/sources.list.bak

geniuspatrick

import numpy as np
import mindspore as ms
from mindspore import nn, ops


class Fold(nn.Cell):
    def __init__(self, channels, output_size, kernel_size, dilation=1, padding=0, stride=1) -> None:
        """Alternative implementation of fold layer via transposed convolution.

        All parameters are the same as `"torch.nn.Fold" <https://pytorch.org/docs/stable/generated/torch.nn.Fold.html>`_,

geniuspatrick

from collections import OrderedDict
import torch
from mindspore import Tensor, Parameter, save_checkpoint
from mindcv.models import resnet50


model = resnet50()
model_weights_ms = model.parameters_dict()
model_weights_pt = torch.load("./resnet50.pth")

geniuspatrick

# refer to https://github.com/rwightman/pytorch-image-models/blob/master/train.py
import yaml
import argparse


# The first arg parser parses out only the --config argument, this argument is used to
# load a yaml file containing key-values that override the defaults for the main parser below
parser_config = argparse.ArgumentParser(description='Training Config', add_help=False)
parser_config.add_argument('-c', '--config', default='', type=str, metavar='FILE',
                           help='YAML config file specifying default arguments')