lint mlperf model_train (#10038)

.github/workflows/test.yml

@@ -338,6 +338,7 @@ jobs:
       run: |
         pip3 install --upgrade --force-reinstall ruff==0.11.0
         python3 -m ruff check .
+        python3 -m ruff check examples/mlperf/model_train.py --ignore E501
     - name: Lint tinygrad with pylint
       run: python -m pylint tinygrad/
     - name: Run mypy

examples/mlperf/dataloader.py

@@ -532,7 +532,7 @@ if __name__ == "__main__":
   def load_retinanet(val):
     from extra.datasets.openimages import BASEDIR, download_dataset
     from pycocotools.coco import COCO
-    dataset = COCO(download_dataset(base_dir:=getenv("BASE_DIR", BASEDIR), "validation" if val else "train"))
+    dataset = COCO(download_dataset(base_dir:=getenv("BASEDIR", BASEDIR), "validation" if val else "train"))
     with tqdm(total=len(dataset.imgs.keys())) as pbar:
       for x in batch_load_retinanet(dataset, val, base_dir):
         pbar.update(x[0].shape[0])

examples/mlperf/model_train.py

@@ -347,12 +347,11 @@ def train_retinanet():
   from examples.mlperf.dataloader import batch_load_retinanet
   from examples.mlperf.initializers import FrozenBatchNorm2dRetinaNet, Conv2dNormalRetinaNet, Conv2dKaimingUniformRetinaNet, Linear, Conv2dRetinaNet
   from extra.datasets.openimages import MLPERF_CLASSES, BASEDIR, download_dataset, normalize, get_dataset_count
-  from extra.models import resnet
+  from extra.models import resnet, retinanet
   from pycocotools.coco import COCO
   from pycocotools.cocoeval import COCOeval
   from tinygrad.helpers import colored
   from typing import Iterator
-  import extra.models.retinanet as retinanet
 
   import numpy as np
 
@@ -399,7 +398,7 @@ def train_retinanet():
     optim.zero_grad()
 
     losses = model(normalize(x, GPUS), **kwargs)
-    loss = sum([l for l in losses.values()])
+    loss = sum(losses.values())
 
     (loss * loss_scaler).backward()
     for t in optim.params: t.grad = t.grad / loss_scaler
@@ -826,7 +825,7 @@ def train_unet3d():
 
         if mean_dice >= TARGET_METRIC:
           is_successful = True
-          save_checkpoint(get_state_dict(model), f"./ckpts/unet3d.safe")
+          save_checkpoint(get_state_dict(model), "./ckpts/unet3d.safe")
         elif mean_dice < 1e-6:
           print("Model diverging. Aborting.")
           diverged = True
@@ -913,7 +912,7 @@ def train_bert():
     MLLOGGER.logger.propagate = False
 
     if INITMLPERF:
-      assert BENCHMARK, f"BENCHMARK must be set for INITMLPERF"
+      assert BENCHMARK, "BENCHMARK must be set for INITMLPERF"
       MLLOGGER.event(key=mllog_constants.SUBMISSION_ORG, value="tinycorp")
       MLLOGGER.event(key=mllog_constants.SUBMISSION_PLATFORM, value=getenv("SUBMISSION_PLATFORM", "tinybox"))
       MLLOGGER.event(key=mllog_constants.SUBMISSION_DIVISION, value=mllog_constants.CLOSED)
@@ -1187,7 +1186,7 @@ def train_bert():
       if MLLOGGER and RUNMLPERF:
         if previous_step:
           MLLOGGER.end(key=mllog_constants.BLOCK_STOP, value=None, metadata={"first_epoch_num": 1, "epoch_num": 1, "first_step_num": i, "step_num": i, "step_count": i - previous_step})
-        MLLOGGER.start(key="checkpoint_start", value=None, metadata={"step_num" : i})
+        MLLOGGER.start(key="checkpoint_start", value=None, metadata={"step_num": i})
       if not os.path.exists(ckpt_dir := save_ckpt_dir): os.mkdir(ckpt_dir)
       if WANDB and wandb.run is not None:
         fn = f"{ckpt_dir}/{time.strftime('%Y%m%d_%H%M%S')}_{wandb.run.id}.safe"

extra/datasets/openimages.py

@@ -205,5 +205,5 @@ def get_dataset_count(base_dir:Path, val:bool) -> int:
   return len(files)
 
 if __name__ == "__main__":
-  download_dataset(base_dir:=getenv("BASE_DIR", BASEDIR), "train")
+  download_dataset(base_dir:=getenv("BASEDIR", BASEDIR), "train")
   download_dataset(base_dir, "validation")