Merge ff1a657f57 into c9ebe86d03

configs/eval.yaml

@@ -0,0 +1,8 @@
+model: 'tf_efficientnet_b0'   # model architecture (default: dpn92)
+img_size: 224                 # Input image dimension
+mean: null                    # Override mean pixel value of dataset
+std: null                     # Override std deviation of of dataset
+num_classes: 2                # Number classes in dataset
+checkpoint: 'output/train/20201124-182940-tf_efficientnet_b0-224/model_best.pth.tar'   # path to latest checkpoint (default: none)
+pretrained: False             # use pre-trained model
+num_gpu: 1                    # Number of GPUS to use

configs/inference.yaml

@@ -0,0 +1,16 @@
+data: 'dataset/test'          # path to dataset
+output_dir: 'output/'         # path to output files
+model: 'tf_efficientnet_b5'   # model architecture (default: dpn92)
+workers: 4                    # number of McD loading workers (default: 2
+batch_size: 256               # mini-batch size (default: 256)
+img_size: 224                 # Input image dimension
+mean: null                    # Override mean pixel value of dataset
+std: null                     # Override std deviation of of dataset
+interpolation: ''             # Image resize interpolation type (overrides model)
+num_classes: 1000             # Number classes in dataset
+log_freq: 10                  # batch logging frequency (default: 10)
+checkpoint: ''                # path to latest checkpoint (default: none)
+pretrained: True              # use pre-trained model
+num_gpu: 1                    # Number of GPUS to use
+no_test_pool: False           # disable test time pool
+topk: 5                       # Top-k to output to CSV

configs/train.yaml

@@ -0,0 +1,102 @@
+# Dataset / Model parameters
+data: 'dataset/splitted'  # path to dataset
+model: 'tf_efficientnet_b0'      # Name of model to train (default: "countception")
+pretrained: True          # Start with pretrained version of specified network (if avail)
+initial_checkpoint: ''    # Initialize model from this checkpoint (default: none)
+resume: ''                # Resume full model and optimizer state from checkpoint (default: none)
+no_resume_opt: False      # prevent resume of optimizer state when resuming model
+num_classes: 2            # number of label classes (default: 1000)
+gp: null                  # Global pool type, one of (fast, avg, max, avgmax, avgmaxc). Model default if None.
+img_size: 224             # Image patch size (default: None => model default)
+crop_pct: null            # Input image center crop percent (for validation only)
+mean: null                # Override mean pixel value of dataset
+std: null                 # Override std deviation of of dataset
+interpolation: ''         # Image resize interpolation type (overrides model)
+batch_size: 16            # input batch size for training (default: 32)
+validation_batch_size_multiplier: 1 # ratio of validation batch size to training batch size (default: 1)
+
+# Optimizer parameters
+opt: 'Adam'               # Optimizer (default: "sgd"
+opt_eps: null             # Optimizer Epsilon (default: None, use opt default)
+opt_betas: null           # Optimizer Betas (default: None, use opt default)
+momentum: 0.9             # Optimizer momentum (default: 0.9)
+weight_decay: 0.0         # weight decay (default: 0.0001)
+clip_grad: null           # Clip gradient norm (default: None, no clipping)
+
+# Learning rate schedule parameters
+sched: 'plateau'          # LR scheduler (default: "step")
+lr: 0.0001                # learning rate (default: 0.01)
+lr_noise: null            # learning rate noise on/off epoch percentages
+lr_noise_pct: 0.67        # learning rate noise limit percent (default: 0.67)
+lr_noise_std: 1.0         # learning rate noise std-dev (default: 1.0)
+lr_cycle_mul: 1.0         # learning rate cycle len multiplier (default: 1.0)
+lr_cycle_limit: 1         # learning rate cycle limit
+warmup_lr: 0.0001         # warmup learning rate (default: 0.0001)
+min_lr: 0.00001           # lower lr bound for cyclic schedulers that hit 0 (1e-5)
+epochs: 30                # number of epochs to train (default: 2)
+start_epoch: null         # manual epoch number (useful on restarts)
+decay_epochs: 5           # epoch interval to decay LR
+warmup_epochs: 10         # epochs to warmup LR, if scheduler supports
+cooldown_epochs: 0        # epochs to cooldown LR at min_lr, after cyclic schedule ends
+patience_epochs: 5        # patience epochs for Plateau LR scheduler (default: 10)
+decay_rate: 0.1           # LR decay rate (default: 0.1)
+
+# Augmentation & regularization parameters
+no_aug: False            # Disable all training augmentation, override other train aug args
+scale: [1, 1]            # Random resize scale (default: 0.08 1.0)
+ratio: [0.8, 1.2]        # Random resize aspect ratio (default: 0.75 1.33)
+hflip: 0.5               # Horizontal flip training aug probability
+vflip: 0.0               # Vertical flip training aug probability
+color_jitter: 0.1        # Color jitter factor (default: 0.4)
+aa: null                 # Use AutoAugment policy. "v0" or "original". (default: None)
+aug_splits: 0            # Number of augmentation splits (default: 0, valid: 0 or >=2)
+jsd: False               # Enable Jensen-Shannon Divergence + CE loss. Use with `--aug-splits`.
+reprob: 0.0              # Random erase prob (default: 0.)
+remode: 'const'          # Random erase mode (default: "const")
+recount: 1               # Random erase count (default: 1)
+resplit: False           # Do not random erase first (clean) augmentation split
+mixup: 0.0               # mixup alpha, mixup enabled if > 0. (default: 0.)
+cutmix: 0.0              # cutmix alpha, cutmix enabled if > 0. (default: 0.
+cutmix_minmax:           # cutmix min/max ratio, overrides alpha and enables cutmix if set (default: None)
+mixup_prob: 1.0          # Probability of performing mixup or cutmix when either/both is enabled
+mixup_switch_prob: 0.5   # Probability of switching to cutmix when both mixup and cutmix enabled
+mixup_mode: 'batch'      # How to apply mixup/cutmix params. Per "batch", "pair", or "elem"
+mixup_off_epoch: 0       # Turn off mixup after this epoch, disabled if 0 (default: 0)
+smoothing: 0.0           # Label smoothing (default: 0.1)
+train_interpolation: 'random' # Training interpolation (random, bilinear, bicubic default: "random"
+drop: 0.0                # Dropout rate (default: 0.)
+drop_connect: null       # Drop connect rate, DEPRECATED, use drop-path (default: None)
+drop_path: null          # Drop path rate (default: None)
+drop_block: null         # Drop block rate (default: None)
+
+# Batch norm parameters (only works with gen_efficientnet based models currently)
+bn_tf: bull              # Use Tensorflow BatchNorm defaults for models that support it (default: False)
+bn_momentum: null        # BatchNorm momentum override (if not None)
+bn_eps: null             # BatchNorm epsilon override (if not None)
+sync_bn: False           # Enable NVIDIA Apex or Torch synchronized BatchNorm.
+dist_bn: ''              # Distribute BatchNorm stats between nodes after each epoch ("broadcast", "reduce", or "")
+split_bn: False          # Enable separate BN layers per augmentation split.
+
+# Model Exponential Moving Average
+model_ema: False         # Enable tracking moving average of model weights
+model_ema_force_cpu: False # Force ema to be tracked on CPU, rank=0 node only. Disables EMA validation.
+model_ema_decay: 0.9998  # decay factor for model weights moving average (default: 0.9998)
+
+# Misc
+seed: 42                 # random seed (default: 42)
+log_interval: 50         # how many batches to wait before logging training status
+recovery_interval: 0     # how many batches to wait before writing recovery checkpoint
+workers: 1               # how many training processes to use (default: 1)
+num_gpu: 1               # Number of GPUS to use
+save_images: False       # save images of input bathes every log interval for debugging
+amp: False               # use NVIDIA Apex AMP or Native AMP for mixed precision training
+apex_amp: False          # Use NVIDIA Apex AMP mixed precision
+native_amp: False        # Use Native Torch AMP mixed precision
+channels_last: False     # Use channels_last memory layout
+pin_mem: False           # Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.
+no_prefetcher: False     # disable fast prefetcher
+output: ''               # path to output folder (default: none, current dir)
+eval_metric: 'top1'      # Best metric (default: "top1"
+tta: 0                   # Test/inference time augmentation (oversampling) factor. 0=None (default: 0)
+local_rank: 0            #
+use_multi_epochs_loader: False # use the multi-epochs-loader to save time at the beginning of every epoch

configs/validate.yaml

@@ -0,0 +1,34 @@
+data: 'dataset/splitted/val'       # path to dataset
+model: 'tf_efficientnet_b0'      # Name of model to train (default: "countception"
+# path to latest checkpoint (default: none)
+checkpoint: 'output/train/tf_efficientnet_b0-224/model_best.pth.tar'
+
+workers: 4                # number of McD loading workers (default: 2)
+batch_size: 16            # mini-batch size (default: 256)
+img_size: 224             # Input image dimension, uses model default if empty
+crop_pct: null            # Input image center crop pct
+mean: null                # Override mean pixel value of dataset
+std: null                 # Override std deviation of of dataset
+interpolation: ''         # Image resize interpolation type (overrides model)
+num_classes: 2            # Number classes in dataset
+class_map: ''             # path to class to idx mapping file (default: "")
+gp: null                  # Global pool type, one of (fast, avg, max, avgmax, avgmaxc). Model default if None.
+log_freq: 10              # batch logging frequency (default: 10)
+pretrained: False         # use pre-trained model
+num_gpu: 1                # Number of GPUS to use
+no_test_pool: False       # disable test time pool
+no_prefetcher: False      # disable fast prefetcher
+pin_mem: False            # Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU
+channels_last: False      # Use channels_last memory layout
+amp: False                # Use AMP mixed precision. Defaults to Apex, fallback to native Torch AMP.
+apex_amp: False           # Use NVIDIA Apex AMP mixed precision
+native_amp: False         # Use Native Torch AMP mixed precision
+tf_preprocessing: False   # Use Tensorflow preprocessing pipeline (require CPU TF installed
+use_ema: False            # use ema version of weights if present
+torchscript: False        # convert model torchscript for inference
+lagacy_jit: False         # use legacy jit mode for pytorch 1.5/1.5.1/1.6 to get back fusion performance
+results_file: ''          # Output csv file for validation results (summary)
+real_labels: ''           # Real labels JSON file for imagenet evaluation
+valid_labels: ''          # Valid label indices txt file for validation of partial label space
+
+

eval.py

@@ -0,0 +1,105 @@
+#!/usr/bin/env python
+"""PyTorch Evaluation Script
+
+An example evaluation script that outputs results of model evaluation.
+
+Hacked together by / Copyright 2020 Ross Wightman (https://github.com/rwightman)
+
+--- Usage: ---
+
+model = ClassificationModel()
+img = Image.open("image.jpg")
+out = model.eval(img)
+print(out)
+
+"""
+import yaml
+from fire import Fire
+from addict import Dict
+
+import torch
+from torchvision import transforms
+
+from timm.models import create_model
+from timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
+
+torch.backends.cudnn.benchmark = True
+
+
+def _update_config(config, params):
+    for k, v in params.items():
+        *path, key = k.split(".")
+        config.update({k: v})
+        print(f"Overwriting {k} = {v} (was {config.get(key)})")
+    return config
+
+
+def _fit(**kwargs):
+    with open('configs/eval.yaml') as stream:
+        base_config = yaml.safe_load(stream)
+
+    if "config" in kwargs.keys():
+        cfg_path = kwargs["config"]
+        with open(cfg_path) as cfg:
+            cfg_yaml = yaml.load(cfg, Loader=yaml.FullLoader)
+
+        merged_cfg = _update_config(base_config, cfg_yaml)
+    else:
+        merged_cfg = base_config
+
+    update_cfg = _update_config(merged_cfg, kwargs)
+    return update_cfg
+
+
+def _parse_args():
+    args = Dict(Fire(_fit))
+
+    # Cache the args as a text string to save them in the output dir later
+    args_text = yaml.safe_dump(args.__dict__, default_flow_style=False)
+    return args, args_text
+
+
+class ClassificationModel:
+    def __init__(self):
+        self.args, self.args_text = _parse_args()
+
+        # might as well try to do something useful...
+        self.args.pretrained = self.args.pretrained or not self.args.checkpoint
+
+        # create model
+        self.model = create_model(
+            self.args.model,
+            num_classes=self.args.num_classes,
+            in_chans=3,
+            pretrained=self.args.pretrained,
+            checkpoint_path=self.args.checkpoint)
+        self.softmax = torch.nn.Softmax(dim=1)
+
+        mean = self.args.mean if self.args.mean is not None else IMAGENET_DEFAULT_MEAN
+        std = self.args.std if self.args.std is not None else IMAGENET_DEFAULT_STD
+        self.loader = transforms.Compose([
+            transforms.ToTensor(),
+            transforms.Resize(self.args.img_size),
+            transforms.Normalize(
+                mean=torch.tensor(mean),
+                std=torch.tensor(std)),
+        ])
+
+        if self.args.num_gpu > 1:
+            self.model = torch.nn.DataParallel(self.model, device_ids=list(range(self.args.num_gpu))).cuda()
+        else:
+            self.model = self.model.cuda()
+            # self.model = self.model.cpu()
+        self.model.eval()
+
+    def eval(self, input):
+        with torch.no_grad():
+            # for OpenCV input
+            # input = Image.fromarray(np.uint8(input)).convert('RGB')
+            input = self.loader(input).float()
+            input = input.cuda()
+
+            labels = self.model(input[None, ...])
+            labels = self.softmax(labels)
+            labels = labels.cpu()
+            return labels.numpy()

inference.py

@@ -7,8 +7,11 @@ Hacked together by / Copyright 2020 Ross Wightman (https://github.com/rwightman)
 """
 import os
 import time
-import argparse
 import logging
+import yaml
+from fire import Fire
+from addict import Dict
+
 import numpy as np
 import torch
 
@@ -20,44 +23,42 @@ torch.backends.cudnn.benchmark = True
 _logger = logging.getLogger('inference')
 
 
-parser = argparse.ArgumentParser(description='PyTorch ImageNet Inference')
-parser.add_argument('data', metavar='DIR',
-                    help='path to dataset')
-parser.add_argument('--output_dir', metavar='DIR', default='./',
-                    help='path to output files')
-parser.add_argument('--model', '-m', metavar='MODEL', default='dpn92',
-                    help='model architecture (default: dpn92)')
-parser.add_argument('-j', '--workers', default=2, type=int, metavar='N',
-                    help='number of data loading workers (default: 2)')
-parser.add_argument('-b', '--batch-size', default=256, type=int,
-                    metavar='N', help='mini-batch size (default: 256)')
-parser.add_argument('--img-size', default=None, type=int,
-                    metavar='N', help='Input image dimension')
-parser.add_argument('--mean', type=float, nargs='+', default=None, metavar='MEAN',
-                    help='Override mean pixel value of dataset')
-parser.add_argument('--std', type=float, nargs='+', default=None, metavar='STD',
-                    help='Override std deviation of of dataset')
-parser.add_argument('--interpolation', default='', type=str, metavar='NAME',
-                    help='Image resize interpolation type (overrides model)')
-parser.add_argument('--num-classes', type=int, default=1000,
-                    help='Number classes in dataset')
-parser.add_argument('--log-freq', default=10, type=int,
-                    metavar='N', help='batch logging frequency (default: 10)')
-parser.add_argument('--checkpoint', default='', type=str, metavar='PATH',
-                    help='path to latest checkpoint (default: none)')
-parser.add_argument('--pretrained', dest='pretrained', action='store_true',
-                    help='use pre-trained model')
-parser.add_argument('--num-gpu', type=int, default=1,
-                    help='Number of GPUS to use')
-parser.add_argument('--no-test-pool', dest='no_test_pool', action='store_true',
-                    help='disable test time pool')
-parser.add_argument('--topk', default=5, type=int,
-                    metavar='N', help='Top-k to output to CSV')
+def _update_config(config, params):
+    for k, v in params.items():
+        *path, key = k.split(".")
+        config.update({k: v})
+        print(f"Overwriting {k} = {v} (was {config.get(key)})")
+    return config
+
+
+def _fit(**kwargs):
+    with open('configs/inference.yaml') as stream:
+        base_config = yaml.safe_load(stream)
+
+    if "config" in kwargs.keys():
+        cfg_path = kwargs["config"]
+        with open(cfg_path) as cfg:
+            cfg_yaml = yaml.load(cfg, Loader=yaml.FullLoader)
+
+        merged_cfg = _update_config(base_config, cfg_yaml)
+    else:
+        merged_cfg = base_config
+
+    update_cfg = _update_config(merged_cfg, kwargs)
+    return update_cfg
+
+
+def _parse_args():
+    args = Dict(Fire(_fit))
+
+    # Cache the args as a text string to save them in the output dir later
+    args_text = yaml.safe_dump(args.__dict__, default_flow_style=False)
+    return args, args_text
 
 
 def main():
     setup_default_logging()
-    args = parser.parse_args()
+    args, args_text = _parse_args()
     # might as well try to do something useful...
     args.pretrained = args.pretrained or not args.checkpoint
 

requirements.txt

@@ -1,3 +1,5 @@
 torch>=1.4.0
 torchvision>=0.5.0
 pyyaml
+fire
+addict

train.py

@@ -14,7 +14,6 @@ NVIDIA CUDA specific speedups adopted from NVIDIA Apex examples
 
 Hacked together by / Copyright 2020 Ross Wightman (https://github.com/rwightman)
 """
-import argparse
 import time
 import yaml
 import os
@@ -22,6 +21,10 @@ import logging
 from collections import OrderedDict
 from contextlib import suppress
 from datetime import datetime
+from fire import Fire
+from addict import Dict
+import numpy as np
+import random
 
 import torch
 import torch.nn as nn
@@ -54,229 +57,54 @@ except AttributeError:
 torch.backends.cudnn.benchmark = True
 _logger = logging.getLogger('train')
 
-# 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
-config_parser = parser = argparse.ArgumentParser(description='Training Config', add_help=False)
-parser.add_argument('-c', '--config', default='', type=str, metavar='FILE',
-                    help='YAML config file specifying default arguments')
-
-
-parser = argparse.ArgumentParser(description='PyTorch ImageNet Training')
-
-# Dataset / Model parameters
-parser.add_argument('data', metavar='DIR',
-                    help='path to dataset')
-parser.add_argument('--model', default='resnet101', type=str, metavar='MODEL',
-                    help='Name of model to train (default: "countception"')
-parser.add_argument('--pretrained', action='store_true', default=False,
-                    help='Start with pretrained version of specified network (if avail)')
-parser.add_argument('--initial-checkpoint', default='', type=str, metavar='PATH',
-                    help='Initialize model from this checkpoint (default: none)')
-parser.add_argument('--resume', default='', type=str, metavar='PATH',
-                    help='Resume full model and optimizer state from checkpoint (default: none)')
-parser.add_argument('--no-resume-opt', action='store_true', default=False,
-                    help='prevent resume of optimizer state when resuming model')
-parser.add_argument('--num-classes', type=int, default=1000, metavar='N',
-                    help='number of label classes (default: 1000)')
-parser.add_argument('--gp', default=None, type=str, metavar='POOL',
-                    help='Global pool type, one of (fast, avg, max, avgmax, avgmaxc). Model default if None.')
-parser.add_argument('--img-size', type=int, default=None, metavar='N',
-                    help='Image patch size (default: None => model default)')
-parser.add_argument('--crop-pct', default=None, type=float,
-                    metavar='N', help='Input image center crop percent (for validation only)')
-parser.add_argument('--mean', type=float, nargs='+', default=None, metavar='MEAN',
-                    help='Override mean pixel value of dataset')
-parser.add_argument('--std', type=float, nargs='+', default=None, metavar='STD',
-                    help='Override std deviation of of dataset')
-parser.add_argument('--interpolation', default='', type=str, metavar='NAME',
-                    help='Image resize interpolation type (overrides model)')
-parser.add_argument('-b', '--batch-size', type=int, default=32, metavar='N',
-                    help='input batch size for training (default: 32)')
-parser.add_argument('-vb', '--validation-batch-size-multiplier', type=int, default=1, metavar='N',
-                    help='ratio of validation batch size to training batch size (default: 1)')
-
-# Optimizer parameters
-parser.add_argument('--opt', default='sgd', type=str, metavar='OPTIMIZER',
-                    help='Optimizer (default: "sgd"')
-parser.add_argument('--opt-eps', default=None, type=float, metavar='EPSILON',
-                    help='Optimizer Epsilon (default: None, use opt default)')
-parser.add_argument('--opt-betas', default=None, type=float, nargs='+', metavar='BETA',
-                    help='Optimizer Betas (default: None, use opt default)')
-parser.add_argument('--momentum', type=float, default=0.9, metavar='M',
-                    help='Optimizer momentum (default: 0.9)')
-parser.add_argument('--weight-decay', type=float, default=0.0001,
-                    help='weight decay (default: 0.0001)')
-parser.add_argument('--clip-grad', type=float, default=None, metavar='NORM',
-                    help='Clip gradient norm (default: None, no clipping)')
-
-
-
-# Learning rate schedule parameters
-parser.add_argument('--sched', default='step', type=str, metavar='SCHEDULER',
-                    help='LR scheduler (default: "step"')
-parser.add_argument('--lr', type=float, default=0.01, metavar='LR',
-                    help='learning rate (default: 0.01)')
-parser.add_argument('--lr-noise', type=float, nargs='+', default=None, metavar='pct, pct',
-                    help='learning rate noise on/off epoch percentages')
-parser.add_argument('--lr-noise-pct', type=float, default=0.67, metavar='PERCENT',
-                    help='learning rate noise limit percent (default: 0.67)')
-parser.add_argument('--lr-noise-std', type=float, default=1.0, metavar='STDDEV',
-                    help='learning rate noise std-dev (default: 1.0)')
-parser.add_argument('--lr-cycle-mul', type=float, default=1.0, metavar='MULT',
-                    help='learning rate cycle len multiplier (default: 1.0)')
-parser.add_argument('--lr-cycle-limit', type=int, default=1, metavar='N',
-                    help='learning rate cycle limit')
-parser.add_argument('--warmup-lr', type=float, default=0.0001, metavar='LR',
-                    help='warmup learning rate (default: 0.0001)')
-parser.add_argument('--min-lr', type=float, default=1e-5, metavar='LR',
-                    help='lower lr bound for cyclic schedulers that hit 0 (1e-5)')
-parser.add_argument('--epochs', type=int, default=200, metavar='N',
-                    help='number of epochs to train (default: 2)')
-parser.add_argument('--start-epoch', default=None, type=int, metavar='N',
-                    help='manual epoch number (useful on restarts)')
-parser.add_argument('--decay-epochs', type=float, default=30, metavar='N',
-                    help='epoch interval to decay LR')
-parser.add_argument('--warmup-epochs', type=int, default=3, metavar='N',
-                    help='epochs to warmup LR, if scheduler supports')
-parser.add_argument('--cooldown-epochs', type=int, default=10, metavar='N',
-                    help='epochs to cooldown LR at min_lr, after cyclic schedule ends')
-parser.add_argument('--patience-epochs', type=int, default=10, metavar='N',
-                    help='patience epochs for Plateau LR scheduler (default: 10')
-parser.add_argument('--decay-rate', '--dr', type=float, default=0.1, metavar='RATE',
-                    help='LR decay rate (default: 0.1)')
-
-# Augmentation & regularization parameters
-parser.add_argument('--no-aug', action='store_true', default=False,
-                    help='Disable all training augmentation, override other train aug args')
-parser.add_argument('--scale', type=float, nargs='+', default=[0.08, 1.0], metavar='PCT',
-                    help='Random resize scale (default: 0.08 1.0)')
-parser.add_argument('--ratio', type=float, nargs='+', default=[3./4., 4./3.], metavar='RATIO',
-                    help='Random resize aspect ratio (default: 0.75 1.33)')
-parser.add_argument('--hflip', type=float, default=0.5,
-                    help='Horizontal flip training aug probability')
-parser.add_argument('--vflip', type=float, default=0.,
-                    help='Vertical flip training aug probability')
-parser.add_argument('--color-jitter', type=float, default=0.4, metavar='PCT',
-                    help='Color jitter factor (default: 0.4)')
-parser.add_argument('--aa', type=str, default=None, metavar='NAME',
-                    help='Use AutoAugment policy. "v0" or "original". (default: None)'),
-parser.add_argument('--aug-splits', type=int, default=0,
-                    help='Number of augmentation splits (default: 0, valid: 0 or >=2)')
-parser.add_argument('--jsd', action='store_true', default=False,
-                    help='Enable Jensen-Shannon Divergence + CE loss. Use with `--aug-splits`.')
-parser.add_argument('--reprob', type=float, default=0., metavar='PCT',
-                    help='Random erase prob (default: 0.)')
-parser.add_argument('--remode', type=str, default='const',
-                    help='Random erase mode (default: "const")')
-parser.add_argument('--recount', type=int, default=1,
-                    help='Random erase count (default: 1)')
-parser.add_argument('--resplit', action='store_true', default=False,
-                    help='Do not random erase first (clean) augmentation split')
-parser.add_argument('--mixup', type=float, default=0.0,
-                    help='mixup alpha, mixup enabled if > 0. (default: 0.)')
-parser.add_argument('--cutmix', type=float, default=0.0,
-                    help='cutmix alpha, cutmix enabled if > 0. (default: 0.)')
-parser.add_argument('--cutmix-minmax', type=float, nargs='+', default=None,
-                    help='cutmix min/max ratio, overrides alpha and enables cutmix if set (default: None)')
-parser.add_argument('--mixup-prob', type=float, default=1.0,
-                    help='Probability of performing mixup or cutmix when either/both is enabled')
-parser.add_argument('--mixup-switch-prob', type=float, default=0.5,
-                    help='Probability of switching to cutmix when both mixup and cutmix enabled')
-parser.add_argument('--mixup-mode', type=str, default='batch',
-                    help='How to apply mixup/cutmix params. Per "batch", "pair", or "elem"')
-parser.add_argument('--mixup-off-epoch', default=0, type=int, metavar='N',
-                    help='Turn off mixup after this epoch, disabled if 0 (default: 0)')
-parser.add_argument('--smoothing', type=float, default=0.1,
-                    help='Label smoothing (default: 0.1)')
-parser.add_argument('--train-interpolation', type=str, default='random',
-                    help='Training interpolation (random, bilinear, bicubic default: "random")')
-parser.add_argument('--drop', type=float, default=0.0, metavar='PCT',
-                    help='Dropout rate (default: 0.)')
-parser.add_argument('--drop-connect', type=float, default=None, metavar='PCT',
-                    help='Drop connect rate, DEPRECATED, use drop-path (default: None)')
-parser.add_argument('--drop-path', type=float, default=None, metavar='PCT',
-                    help='Drop path rate (default: None)')
-parser.add_argument('--drop-block', type=float, default=None, metavar='PCT',
-                    help='Drop block rate (default: None)')
-
-# Batch norm parameters (only works with gen_efficientnet based models currently)
-parser.add_argument('--bn-tf', action='store_true', default=False,
-                    help='Use Tensorflow BatchNorm defaults for models that support it (default: False)')
-parser.add_argument('--bn-momentum', type=float, default=None,
-                    help='BatchNorm momentum override (if not None)')
-parser.add_argument('--bn-eps', type=float, default=None,
-                    help='BatchNorm epsilon override (if not None)')
-parser.add_argument('--sync-bn', action='store_true',
-                    help='Enable NVIDIA Apex or Torch synchronized BatchNorm.')
-parser.add_argument('--dist-bn', type=str, default='',
-                    help='Distribute BatchNorm stats between nodes after each epoch ("broadcast", "reduce", or "")')
-parser.add_argument('--split-bn', action='store_true',
-                    help='Enable separate BN layers per augmentation split.')
-
-# Model Exponential Moving Average
-parser.add_argument('--model-ema', action='store_true', default=False,
-                    help='Enable tracking moving average of model weights')
-parser.add_argument('--model-ema-force-cpu', action='store_true', default=False,
-                    help='Force ema to be tracked on CPU, rank=0 node only. Disables EMA validation.')
-parser.add_argument('--model-ema-decay', type=float, default=0.9998,
-                    help='decay factor for model weights moving average (default: 0.9998)')
-
-# Misc
-parser.add_argument('--seed', type=int, default=42, metavar='S',
-                    help='random seed (default: 42)')
-parser.add_argument('--log-interval', type=int, default=50, metavar='N',
-                    help='how many batches to wait before logging training status')
-parser.add_argument('--recovery-interval', type=int, default=0, metavar='N',
-                    help='how many batches to wait before writing recovery checkpoint')
-parser.add_argument('-j', '--workers', type=int, default=4, metavar='N',
-                    help='how many training processes to use (default: 1)')
-parser.add_argument('--num-gpu', type=int, default=1,
-                    help='Number of GPUS to use')
-parser.add_argument('--save-images', action='store_true', default=False,
-                    help='save images of input bathes every log interval for debugging')
-parser.add_argument('--amp', action='store_true', default=False,
-                    help='use NVIDIA Apex AMP or Native AMP for mixed precision training')
-parser.add_argument('--apex-amp', action='store_true', default=False,
-                    help='Use NVIDIA Apex AMP mixed precision')
-parser.add_argument('--native-amp', action='store_true', default=False,
-                    help='Use Native Torch AMP mixed precision')
-parser.add_argument('--channels-last', action='store_true', default=False,
-                    help='Use channels_last memory layout')
-parser.add_argument('--pin-mem', action='store_true', default=False,
-                    help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')
-parser.add_argument('--no-prefetcher', action='store_true', default=False,
-                    help='disable fast prefetcher')
-parser.add_argument('--output', default='', type=str, metavar='PATH',
-                    help='path to output folder (default: none, current dir)')
-parser.add_argument('--eval-metric', default='top1', type=str, metavar='EVAL_METRIC',
-                    help='Best metric (default: "top1"')
-parser.add_argument('--tta', type=int, default=0, metavar='N',
-                    help='Test/inference time augmentation (oversampling) factor. 0=None (default: 0)')
-parser.add_argument("--local_rank", default=0, type=int)
-parser.add_argument('--use-multi-epochs-loader', action='store_true', default=False,
-                    help='use the multi-epochs-loader to save time at the beginning of every epoch')
 
+def _update_config(config, params):
+    for k, v in params.items():
+        *path, key = k.split(".")
+        config.update({k: v})
+        print(f"Overwriting {k} = {v} (was {config.get(key)})")
+    return config
 
-def _parse_args():
-    # Do we have a config file to parse?
-    args_config, remaining = config_parser.parse_known_args()
-    if args_config.config:
-        with open(args_config.config, 'r') as f:
-            cfg = yaml.safe_load(f)
-            parser.set_defaults(**cfg)
 
-    # The main arg parser parses the rest of the args, the usual
-    # defaults will have been overridden if config file specified.
-    args = parser.parse_args(remaining)
+def _fit(**kwargs):
+    with open('configs/train.yaml') as stream:
+        base_config = yaml.safe_load(stream)
+
+    if "config" in kwargs.keys():
+        cfg_path = kwargs["config"]
+        with open(cfg_path) as cfg:
+            cfg_yaml = yaml.load(cfg, Loader=yaml.FullLoader)
+
+        merged_cfg = _update_config(base_config, cfg_yaml)
+    else:
+        merged_cfg = base_config
+
+    update_cfg = _update_config(merged_cfg, kwargs)
+    return update_cfg
+
+
+def _parse_args():
+    args = Dict(Fire(_fit))
 
     # Cache the args as a text string to save them in the output dir later
     args_text = yaml.safe_dump(args.__dict__, default_flow_style=False)
     return args, args_text
 
 
+def set_deterministic(seed=42, precision=13):
+    np.random.seed(seed)
+    random.seed(seed)
+    # torch.backends.cudnn.benchmarks = False
+    # torch.backends.cudnn.deterministic = True
+    torch.cuda.manual_seed_all(seed)
+    torch.manual_seed(seed)
+    torch.set_printoptions(precision=precision)
+
+
 def main():
     setup_default_logging()
     args, args_text = _parse_args()
+    set_deterministic(args.seed)
 
     args.prefetcher = not args.no_prefetcher
     args.distributed = False

validate.py

@@ -7,17 +7,19 @@ canonical PyTorch, standard Python style, and good performance. Repurpose as you
 
 Hacked together by Ross Wightman (https://github.com/rwightman)
 """
-import argparse
 import os
 import csv
 import glob
 import time
+import yaml
 import logging
 import torch
 import torch.nn as nn
 import torch.nn.parallel
 from collections import OrderedDict
 from contextlib import suppress
+from fire import Fire
+from addict import Dict
 
 from timm.models import create_model, apply_test_time_pool, load_checkpoint, is_model, list_models
 from timm.data import Dataset, DatasetTar, create_loader, resolve_data_config, RealLabelsImagenet
@@ -41,67 +43,37 @@ torch.backends.cudnn.benchmark = True
 _logger = logging.getLogger('validate')
 
 
-parser = argparse.ArgumentParser(description='PyTorch ImageNet Validation')
-parser.add_argument('data', metavar='DIR',
-                    help='path to dataset')
-parser.add_argument('--model', '-m', metavar='MODEL', default='dpn92',
-                    help='model architecture (default: dpn92)')
-parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
-                    help='number of data loading workers (default: 2)')
-parser.add_argument('-b', '--batch-size', default=256, type=int,
-                    metavar='N', help='mini-batch size (default: 256)')
-parser.add_argument('--img-size', default=None, type=int,
-                    metavar='N', help='Input image dimension, uses model default if empty')
-parser.add_argument('--crop-pct', default=None, type=float,
-                    metavar='N', help='Input image center crop pct')
-parser.add_argument('--mean', type=float, nargs='+', default=None, metavar='MEAN',
-                    help='Override mean pixel value of dataset')
-parser.add_argument('--std', type=float,  nargs='+', default=None, metavar='STD',
-                    help='Override std deviation of of dataset')
-parser.add_argument('--interpolation', default='', type=str, metavar='NAME',
-                    help='Image resize interpolation type (overrides model)')
-parser.add_argument('--num-classes', type=int, default=1000,
-                    help='Number classes in dataset')
-parser.add_argument('--class-map', default='', type=str, metavar='FILENAME',
-                    help='path to class to idx mapping file (default: "")')
-parser.add_argument('--gp', default=None, type=str, metavar='POOL',
-                    help='Global pool type, one of (fast, avg, max, avgmax, avgmaxc). Model default if None.')
-parser.add_argument('--log-freq', default=10, type=int,
-                    metavar='N', help='batch logging frequency (default: 10)')
-parser.add_argument('--checkpoint', default='', type=str, metavar='PATH',
-                    help='path to latest checkpoint (default: none)')
-parser.add_argument('--pretrained', dest='pretrained', action='store_true',
-                    help='use pre-trained model')
-parser.add_argument('--num-gpu', type=int, default=1,
-                    help='Number of GPUS to use')
-parser.add_argument('--no-test-pool', dest='no_test_pool', action='store_true',
-                    help='disable test time pool')
-parser.add_argument('--no-prefetcher', action='store_true', default=False,
-                    help='disable fast prefetcher')
-parser.add_argument('--pin-mem', action='store_true', default=False,
-                    help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')
-parser.add_argument('--channels-last', action='store_true', default=False,
-                    help='Use channels_last memory layout')
-parser.add_argument('--amp', action='store_true', default=False,
-                    help='Use AMP mixed precision. Defaults to Apex, fallback to native Torch AMP.')
-parser.add_argument('--apex-amp', action='store_true', default=False,
-                    help='Use NVIDIA Apex AMP mixed precision')
-parser.add_argument('--native-amp', action='store_true', default=False,
-                    help='Use Native Torch AMP mixed precision')
-parser.add_argument('--tf-preprocessing', action='store_true', default=False,
-                    help='Use Tensorflow preprocessing pipeline (require CPU TF installed')
-parser.add_argument('--use-ema', dest='use_ema', action='store_true',
-                    help='use ema version of weights if present')
-parser.add_argument('--torchscript', dest='torchscript', action='store_true',
-                    help='convert model torchscript for inference')
-parser.add_argument('--legacy-jit', dest='legacy_jit', action='store_true',
-                    help='use legacy jit mode for pytorch 1.5/1.5.1/1.6 to get back fusion performance')
-parser.add_argument('--results-file', default='', type=str, metavar='FILENAME',
-                    help='Output csv file for validation results (summary)')
-parser.add_argument('--real-labels', default='', type=str, metavar='FILENAME',
-                    help='Real labels JSON file for imagenet evaluation')
-parser.add_argument('--valid-labels', default='', type=str, metavar='FILENAME',
-                    help='Valid label indices txt file for validation of partial label space')
+def _update_config(config, params):
+    for k, v in params.items():
+        *path, key = k.split(".")
+        config.update({k: v})
+        print(f"Overwriting {k} = {v} (was {config.get(key)})")
+    return config
+
+
+def _fit(**kwargs):
+    with open('configs/validate.yaml') as stream:
+        base_config = yaml.safe_load(stream)
+
+    if "config" in kwargs.keys():
+        cfg_path = kwargs["config"]
+        with open(cfg_path) as cfg:
+            cfg_yaml = yaml.load(cfg, Loader=yaml.FullLoader)
+
+        merged_cfg = _update_config(base_config, cfg_yaml)
+    else:
+        merged_cfg = base_config
+
+    update_cfg = _update_config(merged_cfg, kwargs)
+    return update_cfg
+
+
+def _parse_args():
+    args = Dict(Fire(_fit))
+
+    # Cache the args as a text string to save them in the output dir later
+    args_text = yaml.safe_dump(args.__dict__, default_flow_style=False)
+    return args, args_text
 
 
 def validate(args):
@@ -254,14 +226,15 @@ def validate(args):
         interpolation=data_config['interpolation'])
 
     _logger.info(' * Acc@1 {:.3f} ({:.3f}) Acc@5 {:.3f} ({:.3f})'.format(
-       results['top1'], results['top1_err'], results['top5'], results['top5_err']))
+        results['top1'], results['top1_err'], results['top5'], results['top5_err']))
 
     return results
 
 
 def main():
     setup_default_logging()
-    args = parser.parse_args()
+    args, args_text = _parse_args()
+
     model_cfgs = []
     model_names = []
     if os.path.isdir(args.checkpoint):