mock mnist data for imagenet trainer (#4095)

* mock mnist data for imagenet

* move print and test

* needed to reshape

examples/mlperf/model_train.py

@@ -56,11 +56,8 @@ def train_resnet():
   eval_start_epoch  = getenv("EVAL_START_EPOCH", 0)
   eval_epochs       = getenv("EVAL_EPOCHS", 1)
 
-  if getenv("MOCKDATA"):
-    steps_in_train_epoch, steps_in_val_epoch = 100, 0
-  else:
-    steps_in_train_epoch  = config["steps_in_train_epoch"]  = (len(get_train_files()) // BS)
-    steps_in_val_epoch    = config["steps_in_val_epoch"]    = (len(get_val_files()) // EVAL_BS)
+  steps_in_train_epoch  = config["steps_in_train_epoch"]  = (len(get_train_files()) // BS)
+  steps_in_val_epoch    = config["steps_in_val_epoch"]    = (len(get_val_files()) // EVAL_BS)
 
   config["DEFAULT_FLOAT"] = dtypes.default_float.name
   config["BEAM"]    = BEAM.value
@@ -133,12 +130,7 @@ def train_resnet():
   for e in range(start_epoch, epochs):
     # ** train loop **
     Tensor.training = True
-    if getenv("MOCKDATA"):
-      def mockdata():
-        for _ in range(steps_in_train_epoch): yield Tensor.ones(BS,224,224,3,dtype=dtypes.uint8), [0]*BS, None
-      batch_loader = mockdata()
-    else:
-      batch_loader = batch_load_resnet(batch_size=BS, val=False, shuffle=True, seed=seed*epochs + e)
+    batch_loader = batch_load_resnet(batch_size=BS, val=False, shuffle=True, seed=seed*epochs + e)
     it = iter(tqdm(batch_loader, total=steps_in_train_epoch, desc=f"epoch {e}", disable=BENCHMARK))
     i, proc = 0, data_get(it)
     st = time.perf_counter()

extra/datasets/.gitignore

@@ -1,2 +1,3 @@
 imagenet
 imagenet_bak
+mnist

extra/datasets/fake_imagenet_from_mnist.py

@@ -0,0 +1,41 @@
+#!/usr/bin/env python3
+import pathlib, json
+from tqdm import trange
+from extra.datasets import fetch_mnist
+from PIL import Image
+import numpy as np
+from multiprocessing import Pool
+
+X_train, Y_train, X_test, Y_test = fetch_mnist()
+
+def act(arg):
+  (basedir, i, train) = arg
+  if train:
+    img = np.uint8(X_train[i]).reshape(28, 28)
+    nm = f"train/{Y_train[i]}/{i}.jpg"
+  else:
+    img = np.uint8(X_test[i]).reshape(28, 28)
+    nm = f"val/{Y_test[i]}/{i}.jpg"
+  Image.fromarray(img).resize((224, 224)).convert('RGB').save(basedir / nm)
+
+def create_fake_mnist_imagenet(basedir:pathlib.Path):
+  print(f"creating mock MNIST dataset at {basedir}")
+  basedir.mkdir(exist_ok=True)
+
+  with (basedir / "imagenet_class_index.json").open('w') as f:
+    f.write(json.dumps({str(i):[str(i), str(i)] for i in range(10)}))
+
+  for i in range(10):
+    (basedir / f"train/{i}").mkdir(parents=True, exist_ok=True)
+    (basedir / f"val/{i}").mkdir(parents=True, exist_ok=True)
+
+  def gen(train):
+    for idx in trange(X_train.shape[0] if train else X_test.shape[0]):
+      yield (basedir, idx, train)
+
+  with Pool(64) as p:
+    for _ in p.imap_unordered(act, gen(True)): pass
+    for _ in p.imap_unordered(act, gen(False)): pass
+
+if __name__ == "__main__":
+  create_fake_mnist_imagenet(pathlib.Path("./mnist"))

extra/datasets/imagenet.py

@@ -5,17 +5,29 @@ from PIL import Image
 import functools, pathlib
 from tinygrad.helpers import diskcache, getenv
 
-BASEDIR = pathlib.Path(__file__).parent / "imagenet"
-
 @functools.lru_cache(None)
 def get_imagenet_categories():
   ci = json.load(open(BASEDIR / "imagenet_class_index.json"))
   return {v[0]: int(k) for k,v in ci.items()}
 
-@diskcache
-def get_train_files():
-  if not (files:=glob.glob(p:=str(BASEDIR / "train/*/*"))): raise FileNotFoundError(f"No training files in {p}")
-  return files
+if getenv("MNISTMOCK"):
+  BASEDIR = pathlib.Path(__file__).parent / "mnist"
+
+  @functools.lru_cache(None)
+  def get_train_files():
+    if not BASEDIR.exists():
+      from extra.datasets.fake_imagenet_from_mnist import create_fake_mnist_imagenet
+      create_fake_mnist_imagenet(BASEDIR)
+
+    if not (files:=glob.glob(p:=str(BASEDIR / "train/*/*"))): raise FileNotFoundError(f"No training files in {p}")
+    return files
+else:
+  BASEDIR = pathlib.Path(__file__).parent / "imagenet"
+
+  @diskcache
+  def get_train_files():
+    if not (files:=glob.glob(p:=str(BASEDIR / "train/*/*"))): raise FileNotFoundError(f"No training files in {p}")
+    return files
 
 @functools.lru_cache(None)
 def get_val_files():