mirror of
https://github.com/tinygrad/tinygrad.git
synced 2026-01-26 07:18:40 -05:00
6480a1a180
* add disk_tensor * fix jit * new baseline before whitening * whitening through torch * whiting done currently at 91.65% * 91.99% * clean up mixup and 92.3% * clean up 92.30% * 92.49% before searching for new hyper-parameters * fix CI * fix white space * add whitening init in test * refactor, update hyperpara, 92.72% * converting whiting to tinygrad operation * update CI kernels count for CIFAR * add pad reflect * add random crop 92.53% * update hyperpara 93% * 93.15% on docker container, need to refactor the assignment for hyper param * print out weights and bias to be separated * bias/non-bias params separated * fix whitespace * clean up * refactor hyper-param with dict * refactor lr schedular params * fix whitespace * fix cross entropy loss * fix whitespace * move opt hyp to hyp dict * minor fixup * adjust model, loss scaling * 92.74% while using half of compute as before * update hyp for cutmix * random shuffle during batches * clean up * updating the model * update ConvGroup * disable gradients for batchnorm layer weights * whitespace * 93.92% * clean up * finally 94%git add .! * rewrite whitening to remove dependency on torch * whitespace * remove dependency on torch, 93.91% * back to 94.03% * clean up * update test_real_world
41 lines
2.0 KiB
Python
41 lines
2.0 KiB
Python
import os, random, gzip, tarfile, pickle
|
|
import numpy as np
|
|
from tinygrad.tensor import Tensor
|
|
from tinygrad.helpers import dtypes
|
|
from extra.utils import download_file
|
|
|
|
def fetch_mnist():
|
|
parse = lambda file: np.frombuffer(gzip.open(file).read(), dtype=np.uint8).copy()
|
|
X_train = parse(os.path.dirname(__file__)+"/mnist/train-images-idx3-ubyte.gz")[0x10:].reshape((-1, 28*28)).astype(np.float32)
|
|
Y_train = parse(os.path.dirname(__file__)+"/mnist/train-labels-idx1-ubyte.gz")[8:]
|
|
X_test = parse(os.path.dirname(__file__)+"/mnist/t10k-images-idx3-ubyte.gz")[0x10:].reshape((-1, 28*28)).astype(np.float32)
|
|
Y_test = parse(os.path.dirname(__file__)+"/mnist/t10k-labels-idx1-ubyte.gz")[8:]
|
|
return X_train, Y_train, X_test, Y_test
|
|
|
|
cifar_mean = [0.4913997551666284, 0.48215855929893703, 0.4465309133731618]
|
|
cifar_std = [0.24703225141799082, 0.24348516474564, 0.26158783926049628]
|
|
|
|
def fetch_cifar(shuffle=False):
|
|
def _load_disk_tensor(sz, bs, db_list, path, shuffle=False):
|
|
idx=0
|
|
X, Y = None, None
|
|
for db in db_list:
|
|
x = db[b'data']
|
|
y = np.array(db[b'labels'])
|
|
order = list(range(0, len(y)))
|
|
if shuffle: random.shuffle(order)
|
|
if X is None:
|
|
X = Tensor.empty(sz, *x.shape[1:], device=f'disk:/tmp/{path}'+'_x', dtype=dtypes.uint8)
|
|
Y = Tensor.empty(sz, *y.shape[1:], device=f'disk:/tmp/{path}'+'_y', dtype=dtypes.int64)
|
|
X[idx:idx+bs].assign(x[order,:])
|
|
Y[idx:idx+bs].assign(y[order])
|
|
idx += bs
|
|
return X, Y
|
|
fn = os.path.dirname(__file__)+"/cifar-10-python.tar.gz"
|
|
download_file('https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz', fn)
|
|
tt = tarfile.open(fn, mode='r:gz')
|
|
db = [pickle.load(tt.extractfile(f'cifar-10-batches-py/data_batch_{i}'), encoding="bytes") for i in range(1,6)]
|
|
X_train, Y_train = _load_disk_tensor(50000, 10000, db, "cifar_train", shuffle=shuffle)
|
|
db = [pickle.load(tt.extractfile('cifar-10-batches-py/test_batch'), encoding="bytes")]
|
|
X_test, Y_test = _load_disk_tensor(10000, 10000, db, "cifar_test", shuffle=shuffle)
|
|
return X_train, Y_train, X_test, Y_test |