This hands-on tutorial shows how to fine-tune a Mask R-CNN with a MobileNetV2 backbone using TensorFlow Model Garden. You’ll download LVIS annotations, generate TFRecords from COCO images, load a registered Mask R-CNN experiment config, adapt it for a small custom class set, set training hyperparameters, run distributed training, evaluate, and finally export a SavedModel for deployment. Code and configs follow TensorFlow best practices; original samples are CC BY 4.0 / Apache 2.0.This hands-on tutorial shows how to fine-tune a Mask R-CNN with a MobileNetV2 backbone using TensorFlow Model Garden. You’ll download LVIS annotations, generate TFRecords from COCO images, load a registered Mask R-CNN experiment config, adapt it for a small custom class set, set training hyperparameters, run distributed training, evaluate, and finally export a SavedModel for deployment. Code and configs follow TensorFlow best practices; original samples are CC BY 4.0 / Apache 2.0.
Train a Mask R-CNN for Instance Segmentation with TF Model Garden
Content Overview
- Install Necessary Dependencies
- Import required libraries
- Download subset of Ivis dataset
- Configure the MaskRCNN Resnet FPN COCO model for custom dataset
- Create the Task object (tfm.core.basetask.Task) from the configdefinitions.TaskConfig
\ This tutorial fine-tunes a Mask R-CNN with Mobilenet V2 as backbone model from the TensorFlow Model Garden package (tensorflow-models).
Model Garden contains a collection of state-of-the-art models, implemented with TensorFlow's high-level APIs. The implementations demonstrate the best practices for modeling, letting users to take full advantage of TensorFlow for their research and product development.
This tutorial demonstrates how to:
- Use models from the TensorFlow Models package.
- Train/Fine-tune a pre-built Mask R-CNN with mobilenet as backbone for Object Detection and Instance Segmentation
- Export the trained/tuned Mask R-CNN model
Install Necessary Dependencies
pip install -U -q "tf-models-official" pip install -U -q remotezip tqdm opencv-python einops Import required libraries
import os import io import json import tqdm import shutil import pprint import pathlib import tempfile import requests import collections import matplotlib import numpy as np import tensorflow as tf import matplotlib.pyplot as plt from PIL import Image from six import BytesIO from etils import epath from IPython import display from urllib.request import urlopen \
2023-11-30 12:05:19.630836: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered 2023-11-30 12:05:19.630880: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered 2023-11-30 12:05:19.632442: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered \
import orbit import tensorflow as tf import tensorflow_models as tfm import tensorflow_datasets as tfds from official.core import exp_factory from official.core import config_definitions as cfg from official.vision.data import tfrecord_lib from official.vision.serving import export_saved_model_lib from official.vision.dataloaders.tf_example_decoder import TfExampleDecoder from official.vision.utils.object_detection import visualization_utils from official.vision.ops.preprocess_ops import normalize_image, resize_and_crop_image from official.vision.data.create_coco_tf_record import coco_annotations_to_lists pp = pprint.PrettyPrinter(indent=4) # Set Pretty Print Indentation print(tf.__version__) # Check the version of tensorflow used %matplotlib inline \
2.15.0 Download subset of lvis dataset
LVIS: A dataset for large vocabulary instance segmentation.
\
:::tip Note: LVIS uses the COCO 2017 train, validation, and test image sets. If you have already downloaded the COCO images, you only need to download the LVIS annotations. LVIS val set contains images from COCO 2017 train in addition to the COCO 2017 val split.
:::
\
# @title Download annotation files wget https://dl.fbaipublicfiles.com/LVIS/lvis_v1_train.json.zip unzip -q lvis_v1_train.json.zip rm lvis_v1_train.json.zip wget https://dl.fbaipublicfiles.com/LVIS/lvis_v1_val.json.zip unzip -q lvis_v1_val.json.zip rm lvis_v1_val.json.zip wget https://dl.fbaipublicfiles.com/LVIS/lvis_v1_image_info_test_dev.json.zip unzip -q lvis_v1_image_info_test_dev.json.zip rm lvis_v1_image_info_test_dev.json.zip \
--2023-11-30 12:05:23-- https://dl.fbaipublicfiles.com/LVIS/lvis_v1_train.json.zip Resolving dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)... 3.163.189.51, 3.163.189.108, 3.163.189.14, ... Connecting to dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)|3.163.189.51|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 350264821 (334M) [application/zip] Saving to: ‘lvis_v1_train.json.zip’ lvis_v1_train.json. 100%[===================>] 334.04M 295MB/s in 1.1s 2023-11-30 12:05:25 (295 MB/s) - ‘lvis_v1_train.json.zip’ saved [350264821/350264821] --2023-11-30 12:05:34-- https://dl.fbaipublicfiles.com/LVIS/lvis_v1_val.json.zip Resolving dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)... 3.163.189.51, 3.163.189.108, 3.163.189.14, ... Connecting to dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)|3.163.189.51|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 64026968 (61M) [application/zip] Saving to: ‘lvis_v1_val.json.zip’ lvis_v1_val.json.zi 100%[===================>] 61.06M 184MB/s in 0.3s 2023-11-30 12:05:34 (184 MB/s) - ‘lvis_v1_val.json.zip’ saved [64026968/64026968] --2023-11-30 12:05:36-- https://dl.fbaipublicfiles.com/LVIS/lvis_v1_image_info_test_dev.json.zip Resolving dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)... 3.163.189.51, 3.163.189.108, 3.163.189.14, ... Connecting to dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)|3.163.189.51|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 384629 (376K) [application/zip] Saving to: ‘lvis_v1_image_info_test_dev.json.zip’ lvis_v1_image_info_ 100%[===================>] 375.61K --.-KB/s in 0.03s 2023-11-30 12:05:37 (12.3 MB/s) - ‘lvis_v1_image_info_test_dev.json.zip’ saved [384629/384629] \
# @title Lvis annotation parsing # Annotations with invalid bounding boxes. Will not be used. _INVALID_ANNOTATIONS = [ # Train split. 662101, 81217, 462924, 227817, 29381, 601484, 412185, 504667, 572573, 91937, 239022, 181534, 101685, # Validation split. 36668, 57541, 33126, 10932, ] def get_category_map(annotation_path, num_classes): with epath.Path(annotation_path).open() as f: data = json.load(f) category_map = {id+1: {'id': cat_dict['id'], 'name': cat_dict['name']} for id, cat_dict in enumerate(data['categories'][:num_classes])} return category_map class LvisAnnotation: """LVIS annotation helper class. The format of the annations is explained on https://www.lvisdataset.org/dataset. """ def __init__(self, annotation_path): with epath.Path(annotation_path).open() as f: data = json.load(f) self._data = data img_id2annotations = collections.defaultdict(list) for a in self._data.get('annotations', []): if a['category_id'] in category_ids: img_id2annotations[a['image_id']].append(a) self._img_id2annotations = { k: list(sorted(v, key=lambda a: a['id'])) for k, v in img_id2annotations.items() } @property def categories(self): """Return the category dicts, as sorted in the file.""" return self._data['categories'] @property def images(self): """Return the image dicts, as sorted in the file.""" sub_images = [] for image_info in self._data['images']: if image_info['id'] in self._img_id2annotations: sub_images.append(image_info) return sub_images def get_annotations(self, img_id): """Return all annotations associated with the image id string.""" # Some images don't have any annotations. Return empty list instead. return self._img_id2annotations.get(img_id, []) def _generate_tf_records(prefix, images_zip, annotation_file, num_shards=5): """Generate TFRecords.""" lvis_annotation = LvisAnnotation(annotation_file) def _process_example(prefix, image_info, id_to_name_map): # Search image dirs. filename = pathlib.Path(image_info['coco_url']).name image = tf.io.read_file(os.path.join(IMGS_DIR, filename)) instances = lvis_annotation.get_annotations(img_id=image_info['id']) instances = [x for x in instances if x['id'] not in _INVALID_ANNOTATIONS] # print([x['category_id'] for x in instances]) is_crowd = {'iscrowd': 0} instances = [dict(x, **is_crowd) for x in instances] neg_category_ids = image_info.get('neg_category_ids', []) not_exhaustive_category_ids = image_info.get( 'not_exhaustive_category_ids', [] ) data, _ = coco_annotations_to_lists(instances, id_to_name_map, image_info['height'], image_info['width'], include_masks=True) # data['category_id'] = [id-1 for id in data['category_id']] keys_to_features = { 'image/encoded': tfrecord_lib.convert_to_feature(image.numpy()), 'image/filename': tfrecord_lib.convert_to_feature(filename.encode('utf8')), 'image/format': tfrecord_lib.convert_to_feature('jpg'.encode('utf8')), 'image/height': tfrecord_lib.convert_to_feature(image_info['height']), 'image/width': tfrecord_lib.convert_to_feature(image_info['width']), 'image/source_id': tfrecord_lib.convert_to_feature(str(image_info['id']).encode('utf8')), 'image/object/bbox/xmin': tfrecord_lib.convert_to_feature(data['xmin']), 'image/object/bbox/xmax': tfrecord_lib.convert_to_feature(data['xmax']), 'image/object/bbox/ymin': tfrecord_lib.convert_to_feature(data['ymin']), 'image/object/bbox/ymax': tfrecord_lib.convert_to_feature(data['ymax']), 'image/object/class/text': tfrecord_lib.convert_to_feature(data['category_names']), 'image/object/class/label': tfrecord_lib.convert_to_feature(data['category_id']), 'image/object/is_crowd': tfrecord_lib.convert_to_feature(data['is_crowd']), 'image/object/area': tfrecord_lib.convert_to_feature(data['area'], 'float_list'), 'image/object/mask': tfrecord_lib.convert_to_feature(data['encoded_mask_png']) } # print(keys_to_features['image/object/class/label']) example = tf.train.Example( features=tf.train.Features(feature=keys_to_features)) return example # file_names = [f"{prefix}/{pathlib.Path(image_info['coco_url']).name}" # for image_info in lvis_annotation.images] # _extract_images(images_zip, file_names) writers = [ tf.io.TFRecordWriter( tf_records_dir + prefix +'-%05d-of-%05d.tfrecord' % (i, num_shards)) for i in range(num_shards) ] id_to_name_map = {cat_dict['id']: cat_dict['name'] for cat_dict in lvis_annotation.categories[:NUM_CLASSES]} # print(id_to_name_map) for idx, image_info in enumerate(tqdm.tqdm(lvis_annotation.images)): img_data = requests.get(image_info['coco_url'], stream=True).content img_name = image_info['coco_url'].split('/')[-1] with open(os.path.join(IMGS_DIR, img_name), 'wb') as handler: handler.write(img_data) tf_example = _process_example(prefix, image_info, id_to_name_map) writers[idx % num_shards].write(tf_example.SerializeToString()) del lvis_annotation \
_URLS = { 'train_images': 'http://images.cocodataset.org/zips/train2017.zip', 'validation_images': 'http://images.cocodataset.org/zips/val2017.zip', 'test_images': 'http://images.cocodataset.org/zips/test2017.zip', } train_prefix = 'train' valid_prefix = 'val' train_annotation_path = './lvis_v1_train.json' valid_annotation_path = './lvis_v1_val.json' IMGS_DIR = './lvis_sub_dataset/' tf_records_dir = './lvis_tfrecords/' if not os.path.exists(IMGS_DIR): os.mkdir(IMGS_DIR) if not os.path.exists(tf_records_dir): os.mkdir(tf_records_dir) NUM_CLASSES = 3 category_index = get_category_map(valid_annotation_path, NUM_CLASSES) category_ids = list(category_index.keys()) \
# Below helper function are taken from github tensorflow dataset lvis # https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/datasets/lvis/lvis_dataset_builder.py _generate_tf_records(train_prefix, _URLS['train_images'], train_annotation_path) \
100%|██████████| 2338/2338 [16:14<00:00, 2.40it/s] \
_generate_tf_records(valid_prefix, _URLS['validation_images'], valid_annotation_path) \
100%|██████████| 422/422 [02:56<00:00, 2.40it/s] Configure the MaskRCNN Resnet FPN COCO model for custom dataset
train_data_input_path = './lvis_tfrecords/train*' valid_data_input_path = './lvis_tfrecords/val*' test_data_input_path = './lvis_tfrecords/test*' model_dir = './trained_model/' export_dir ='./exported_model/' \
if not os.path.exists(model_dir): os.mkdir(model_dir) In Model Garden, the collections of parameters that define a model are called configs. Model Garden can create a config based on a known set of parameters via a factory.
Use the retinanet_mobilenet_coco experiment configuration, as defined by tfm.vision.configs.maskrcnn.maskrcnn_mobilenet_coco.
Please find all the registered experiements here
The configuration defines an experiment to train a Mask R-CNN model with mobilenet as backbone and FPN as decoder. Default Congiguration is trained on COCO train2017 and evaluated on COCO val2017.
There are also other alternative experiments available such as maskrcnn_resnetfpn_coco, maskrcnn_spinenet_coco and more. One can switch to them by changing the experiment name argument to the get_exp_config function.
\
exp_config = exp_factory.get_exp_config('maskrcnn_mobilenet_coco') \
model_ckpt_path = './model_ckpt/' if not os.path.exists(model_ckpt_path): os.mkdir(model_ckpt_path) !gsutil cp gs://tf_model_garden/vision/mobilenet/v2_1.0_float/ckpt-180648.data-00000-of-00001 './model_ckpt/' !gsutil cp gs://tf_model_garden/vision/mobilenet/v2_1.0_float/ckpt-180648.index './model_ckpt/' \
Copying gs://tf_model_garden/vision/mobilenet/v2_1.0_float/ckpt-180648.data-00000-of-00001... Operation completed over 1 objects/26.9 MiB. Copying gs://tf_model_garden/vision/mobilenet/v2_1.0_float/ckpt-180648.index... Operation completed over 1 objects/7.5 KiB. Adjust the model and dataset configurations so that it works with custom dataset.
BATCH_SIZE = 8 HEIGHT, WIDTH = 256, 256 IMG_SHAPE = [HEIGHT, WIDTH, 3] # Backbone Config exp_config.task.annotation_file = None exp_config.task.freeze_backbone = True exp_config.task.init_checkpoint = "./model_ckpt/ckpt-180648" exp_config.task.init_checkpoint_modules = "backbone" # Model Config exp_config.task.model.num_classes = NUM_CLASSES + 1 exp_config.task.model.input_size = IMG_SHAPE # Training Data Config exp_config.task.train_data.input_path = train_data_input_path exp_config.task.train_data.dtype = 'float32' exp_config.task.train_data.global_batch_size = BATCH_SIZE exp_config.task.train_data.shuffle_buffer_size = 64 exp_config.task.train_data.parser.aug_scale_max = 1.0 exp_config.task.train_data.parser.aug_scale_min = 1.0 # Validation Data Config exp_config.task.validation_data.input_path = valid_data_input_path exp_config.task.validation_data.dtype = 'float32' exp_config.task.validation_data.global_batch_size = BATCH_SIZE Adjust the trainer configuration.
logical_device_names = [logical_device.name for logical_device in tf.config.list_logical_devices()] if 'GPU' in ''.join(logical_device_names): print('This may be broken in Colab.') device = 'GPU' elif 'TPU' in ''.join(logical_device_names): print('This may be broken in Colab.') device = 'TPU' else: print('Running on CPU is slow, so only train for a few steps.') device = 'CPU' train_steps = 2000 exp_config.trainer.steps_per_loop = 200 # steps_per_loop = num_of_training_examples // train_batch_size exp_config.trainer.summary_interval = 200 exp_config.trainer.checkpoint_interval = 200 exp_config.trainer.validation_interval = 200 exp_config.trainer.validation_steps = 200 # validation_steps = num_of_validation_examples // eval_batch_size exp_config.trainer.train_steps = train_steps exp_config.trainer.optimizer_config.warmup.linear.warmup_steps = 200 exp_config.trainer.optimizer_config.learning_rate.type = 'cosine' exp_config.trainer.optimizer_config.learning_rate.cosine.decay_steps = train_steps exp_config.trainer.optimizer_config.learning_rate.cosine.initial_learning_rate = 0.07 exp_config.trainer.optimizer_config.warmup.linear.warmup_learning_rate = 0.05 \
This may be broken in Colab. Print the modified configuration.
pp.pprint(exp_config.as_dict()) display.Javascript("google.colab.output.setIframeHeight('500px');") \
{ 'runtime': { 'all_reduce_alg': None, 'batchnorm_spatial_persistent': False, 'dataset_num_private_threads': None, 'default_shard_dim': -1, 'distribution_strategy': 'mirrored', 'enable_xla': False, 'gpu_thread_mode': None, 'loss_scale': None, 'mixed_precision_dtype': 'bfloat16', 'num_cores_per_replica': 1, 'num_gpus': 0, 'num_packs': 1, 'per_gpu_thread_count': 0, 'run_eagerly': False, 'task_index': -1, 'tpu': None, 'tpu_enable_xla_dynamic_padder': None, 'use_tpu_mp_strategy': False, 'worker_hosts': None}, 'task': { 'allow_image_summary': False, 'allowed_mask_class_ids': None, 'annotation_file': None, 'differential_privacy_config': None, 'freeze_backbone': True, 'init_checkpoint': './model_ckpt/ckpt-180648', 'init_checkpoint_modules': 'backbone', 'losses': { 'class_weights': None, 'frcnn_box_weight': 1.0, 'frcnn_class_loss_top_k_percent': 1.0, 'frcnn_class_use_binary_cross_entropy': False, 'frcnn_class_weight': 1.0, 'frcnn_huber_loss_delta': 1.0, 'l2_weight_decay': 4e-05, 'loss_weight': 1.0, 'mask_weight': 1.0, 'rpn_box_weight': 1.0, 'rpn_huber_loss_delta': 0.1111111111111111, 'rpn_score_weight': 1.0}, 'model': { 'anchor': { 'anchor_size': 3, 'aspect_ratios': [0.5, 1.0, 2.0], 'num_scales': 1}, 'backbone': { 'mobilenet': { 'filter_size_scale': 1.0, 'model_id': 'MobileNetV2', 'output_intermediate_endpoints': False, 'output_stride': None, 'stochastic_depth_drop_rate': 0.0}, 'type': 'mobilenet'}, 'decoder': { 'fpn': { 'fusion_type': 'sum', 'num_filters': 128, 'use_keras_layer': False, 'use_separable_conv': True}, 'type': 'fpn'}, 'detection_generator': { 'apply_nms': True, 'max_num_detections': 100, 'nms_iou_threshold': 0.5, 'nms_version': 'v2', 'pre_nms_score_threshold': 0.05, 'pre_nms_top_k': 5000, 'soft_nms_sigma': None, 'use_cpu_nms': False, 'use_sigmoid_probability': False}, 'detection_head': { 'cascade_class_ensemble': False, 'class_agnostic_bbox_pred': False, 'fc_dims': 512, 'num_convs': 4, 'num_fcs': 1, 'num_filters': 128, 'use_separable_conv': True}, 'include_mask': True, 'input_size': [256, 256, 3], 'mask_head': { 'class_agnostic': False, 'num_convs': 4, 'num_filters': 128, 'upsample_factor': 2, 'use_separable_conv': True}, 'mask_roi_aligner': { 'crop_size': 14, 'sample_offset': 0.5}, 'mask_sampler': {'num_sampled_masks': 128}, 'max_level': 6, 'min_level': 3, 'norm_activation': { 'activation': 'relu6', 'norm_epsilon': 0.001, 'norm_momentum': 0.99, 'use_sync_bn': True}, 'num_classes': 4, 'outer_boxes_scale': 1.0, 'roi_aligner': { 'crop_size': 7, 'sample_offset': 0.5}, 'roi_generator': { 'nms_iou_threshold': 0.7, 'num_proposals': 1000, 'pre_nms_min_size_threshold': 0.0, 'pre_nms_score_threshold': 0.0, 'pre_nms_top_k': 2000, 'test_nms_iou_threshold': 0.7, 'test_num_proposals': 1000, 'test_pre_nms_min_size_threshold': 0.0, 'test_pre_nms_score_threshold': 0.0, 'test_pre_nms_top_k': 1000, 'use_batched_nms': False}, 'roi_sampler': { 'background_iou_high_threshold': 0.5, 'background_iou_low_threshold': 0.0, 'cascade_iou_thresholds': None, 'foreground_fraction': 0.25, 'foreground_iou_threshold': 0.5, 'mix_gt_boxes': True, 'num_sampled_rois': 512}, 'rpn_head': { 'num_convs': 1, 'num_filters': 128, 'use_separable_conv': True} }, 'name': None, 'per_category_metrics': False, 'train_data': { 'apply_tf_data_service_before_batching': False, 'autotune_algorithm': None, 'block_length': 1, 'cache': False, 'cycle_length': None, 'decoder': { 'simple_decoder': { 'attribute_names': [ ], 'mask_binarize_threshold': None, 'regenerate_source_id': False}, 'type': 'simple_decoder'}, 'deterministic': None, 'drop_remainder': True, 'dtype': 'float32', 'enable_shared_tf_data_service_between_parallel_trainers': False, 'enable_tf_data_service': False, 'file_type': 'tfrecord', 'global_batch_size': 8, 'input_path': './lvis_tfrecords/train*', 'is_training': True, 'num_examples': -1, 'parser': { 'aug_rand_hflip': True, 'aug_rand_vflip': False, 'aug_scale_max': 1.0, 'aug_scale_min': 1.0, 'aug_type': None, 'mask_crop_size': 112, 'match_threshold': 0.5, 'max_num_instances': 100, 'num_channels': 3, 'pad': True, 'rpn_batch_size_per_im': 256, 'rpn_fg_fraction': 0.5, 'rpn_match_threshold': 0.7, 'rpn_unmatched_threshold': 0.3, 'skip_crowd_during_training': True, 'unmatched_threshold': 0.5}, 'prefetch_buffer_size': None, 'seed': None, 'sharding': True, 'shuffle_buffer_size': 64, 'tf_data_service_address': None, 'tf_data_service_job_name': None, 'tfds_as_supervised': False, 'tfds_data_dir': '', 'tfds_name': '', 'tfds_skip_decoding_feature': '', 'tfds_split': '', 'trainer_id': None, 'weights': None}, 'use_approx_instance_metrics': False, 'use_coco_metrics': True, 'use_wod_metrics': False, 'validation_data': { 'apply_tf_data_service_before_batching': False, 'autotune_algorithm': None, 'block_length': 1, 'cache': False, 'cycle_length': None, 'decoder': { 'simple_decoder': { 'attribute_names': [ ], 'mask_binarize_threshold': None, 'regenerate_source_id': False}, 'type': 'simple_decoder'}, 'deterministic': None, 'drop_remainder': False, 'dtype': 'float32', 'enable_shared_tf_data_service_between_parallel_trainers': False, 'enable_tf_data_service': False, 'file_type': 'tfrecord', 'global_batch_size': 8, 'input_path': './lvis_tfrecords/val*', 'is_training': False, 'num_examples': -1, 'parser': { 'aug_rand_hflip': False, 'aug_rand_vflip': False, 'aug_scale_max': 1.0, 'aug_scale_min': 1.0, 'aug_type': None, 'mask_crop_size': 112, 'match_threshold': 0.5, 'max_num_instances': 100, 'num_channels': 3, 'pad': True, 'rpn_batch_size_per_im': 256, 'rpn_fg_fraction': 0.5, 'rpn_match_threshold': 0.7, 'rpn_unmatched_threshold': 0.3, 'skip_crowd_during_training': True, 'unmatched_threshold': 0.5}, 'prefetch_buffer_size': None, 'seed': None, 'sharding': True, 'shuffle_buffer_size': 10000, 'tf_data_service_address': None, 'tf_data_service_job_name': None, 'tfds_as_supervised': False, 'tfds_data_dir': '', 'tfds_name': '', 'tfds_skip_decoding_feature': '', 'tfds_split': '', 'trainer_id': None, 'weights': None} }, 'trainer': { 'allow_tpu_summary': False, 'best_checkpoint_eval_metric': '', 'best_checkpoint_export_subdir': '', 'best_checkpoint_metric_comp': 'higher', 'checkpoint_interval': 200, 'continuous_eval_timeout': 3600, 'eval_tf_function': True, 'eval_tf_while_loop': False, 'loss_upper_bound': 1000000.0, 'max_to_keep': 5, 'optimizer_config': { 'ema': None, 'learning_rate': { 'cosine': { 'alpha': 0.0, 'decay_steps': 2000, 'initial_learning_rate': 0.07, 'name': 'CosineDecay', 'offset': 0}, 'type': 'cosine'}, 'optimizer': { 'sgd': { 'clipnorm': None, 'clipvalue': None, 'decay': 0.0, 'global_clipnorm': None, 'momentum': 0.9, 'name': 'SGD', 'nesterov': False}, 'type': 'sgd'}, 'warmup': { 'linear': { 'name': 'linear', 'warmup_learning_rate': 0.05, 'warmup_steps': 200}, 'type': 'linear'} }, 'preemption_on_demand_checkpoint': True, 'recovery_begin_steps': 0, 'recovery_max_trials': 0, 'steps_per_loop': 200, 'summary_interval': 200, 'train_steps': 2000, 'train_tf_function': True, 'train_tf_while_loop': True, 'validation_interval': 200, 'validation_steps': 200, 'validation_summary_subdir': 'validation'} } <IPython.core.display.Javascript object> Set up the distribution strategy.
# Setting up the Strategy if exp_config.runtime.mixed_precision_dtype == tf.float16: tf.keras.mixed_precision.set_global_policy('mixed_float16') if 'GPU' in ''.join(logical_device_names): distribution_strategy = tf.distribute.MirroredStrategy() elif 'TPU' in ''.join(logical_device_names): tf.tpu.experimental.initialize_tpu_system() tpu = tf.distribute.cluster_resolver.TPUClusterResolver(tpu='/device:TPU_SYSTEM:0') distribution_strategy = tf.distribute.experimental.TPUStrategy(tpu) else: print('Warning: this will be really slow.') distribution_strategy = tf.distribute.OneDeviceStrategy(logical_device_names[0]) print("Done") \
INFO:tensorflow:Using MirroredStrategy with devices ('/job:localhost/replica:0/task:0/device:GPU:0', '/job:localhost/replica:0/task:0/device:GPU:1', '/job:localhost/replica:0/task:0/device:GPU:2', '/job:localhost/replica:0/task:0/device:GPU:3') Done Create the Task object (tfm.core.base_task.Task) from the config_definitions.TaskConfig.
The Task object has all the methods necessary for building the dataset, building the model, and running training & evaluation. These methods are driven by tfm.core.train_lib.run_experiment.
\
with distribution_strategy.scope(): task = tfm.core.task_factory.get_task(exp_config.task, logging_dir=model_dir) \ \
:::info Originally published on the TensorFlow website, this article appears here under a new headline and is licensed under CC BY 4.0. Code samples shared under the Apache 2.0 License.
:::
\
Disclaimer: The articles reposted on this site are sourced from public platforms and are provided for informational purposes only. They do not necessarily reflect the views of MEXC. All rights remain with the original authors. If you believe any content infringes on third-party rights, please contact service@support.mexc.com for removal. MEXC makes no guarantees regarding the accuracy, completeness, or timeliness of the content and is not responsible for any actions taken based on the information provided. The content does not constitute financial, legal, or other professional advice, nor should it be considered a recommendation or endorsement by MEXC.
You May Also Like
Vitalik Buterin Proposes Onchain Gas Futures to Stabilize Ethereum Fees
Vitalik Buterin said that the Ethereum gas futures market would allow developers, traders, and institutions to hedge against fee volatility. The proposal comes as Ethereum prepares for major scaling upgrades, including a 5x increase in the gas limit. Vitalik Buterin, the co-founder of Ethereum blockchain, shared a new idea on having an on-chain futures market [...]]]>
Share
Crypto News Flash2025/12/08 15:43
Trading Moment: Markets Enter a Key Week Ending the Year, Bitcoin Holds Key Level at $86,000
Daily market data review and trend analysis, produced by PANews. 1. Market Observation Markets are holding their breath for this week's Federal Reserve meeting, with a 25-basis-point rate cut widely expected. However, contrary to conventional wisdom, since the rate-cutting cycle began in September, the yield on long-term US Treasury bonds, the anchor for global asset pricing, has risen instead of falling, triggering intense debate about the future economic path. Optimists see this as a signal of a "soft landing," while pessimists worry it's a vote of no confidence from the "bond vigilantes" regarding the high national debt and inflation risks in the US. Against this backdrop, Wall Street veteran strategists like Mark Cabana of Bank of America predict that, in addition to rate cuts, the Fed may announce a major balance sheet expansion plan of up to $45 billion per month to address potential liquidity shortages. Meanwhile, China will also usher in a super week of policy announcements, with important meetings and the release of key economic data such as inflation and social financing providing new guidance for the market. Furthermore, competition in the field of artificial intelligence is becoming increasingly fierce, with OpenAI planning to release GPT-5.2 ahead of schedule to address this competition. The financial reports of Broadcom, a chip designer and Oracle, both core players in the AI industry chain, as well as the visit of Microsoft's CEO to India, will all serve as key indicators for assessing the investment climate in AI infrastructure and the future direction of the industry. In the Bitcoin market, short-term sentiment is cautious, but long-term indicators remain resilient. Analyst Murphy, based on the MVRV indicator, predicts that Bitcoin's price may reach $85,000 to $94,000 by December 31st, and then touch the $71,000 to $104,000 range in early 2026, considering $104,000 as a key bull-bear dividing line. Several analysts consider the $86,000 to $88,000 area as key support. For example, Daan Crypto Trades points out that a break below this key Fibonacci level could lead to a price pullback to a low of $76,000, while Michaël van de Poppe believes that holding $86,000 is a prerequisite for his bullish scenario (i.e., a price break above $92,000 and head towards $100,000). On-chain data presents a mixed picture: on the one hand, Glassnode points out that ETF demand continues to weaken, and market risk appetite is declining; on the other hand, analyst @TXMCtrades emphasizes the continued rise in the "activity" indicator, and CryptoQuant data also shows that selling pressure from long-term holders has been "completely reset," which may indicate potential spot demand and the formation of a market bottom. Bloomberg ETF expert Eric Balchunas, however, offers a more macro-level reassurance to the market, believing that Bitcoin's correction this year is merely a normal cooling down of last year's extreme 122% surge. Its resilience in reaching new highs after multiple significant pullbacks makes it no longer suitable for comparison to the "tulip bubble." Regarding Ethereum, short-term market sentiment leans towards pessimism, but long-term technical patterns are showing optimistic signals. According to Nansen data, "smart money" traders are still adding to their short positions in Ethereum on the derivatives platform Hyperliquid, with net short positions accumulating to over $21 million. However, analyst Sykodelic sees a positive side in the technical charts, pointing out that Ethereum's 5-day MACD and RSI indicators, after a thorough reset, are exhibiting patterns that have historically led to significant rallies, suggesting that a market bottom is forming. In the altcoin market, the AI project Bittensor (TAO) became the focus of attention. The project will undergo its first halving on December 14th, reducing the daily token issuance by half. Grayscale analyst Will Ogden Moore commented positively, believing it marks a significant milestone in the network's maturation. He pointed out that its strong adoption momentum, rising institutional interest, and the success of the dTAO mechanism could all be catalysts for price increases. TAO rose nearly 10% intraday. The weekend saw numerous market developments, with several events and figures attracting widespread attention. Terraform Labs co-founder Do Kwon's legal case saw new developments. US prosecutors recommended a 12-year prison sentence for his "massive" fraudulent activities, and US District Judge Paul Engelmayer will deliver sentencing on December 11th. This news initially caused USTC and LUNA tokens to surge by over 100% over the weekend before falling sharply, down nearly 20% in the past 24 hours. Additionally, Binance founder CZ's joke about executive He Yi's misspelling of "DOYR" in a tweet unexpectedly spawned a meme coin with the same name. Meanwhile, Binance responded directly to community concerns, stating that it is conducting an internal review of potential corruption related to token listings. Another noteworthy piece of news comes from the intersection of the tech and cryptocurrency worlds: Moore Threads, the "first domestically produced GPU stock," saw its share price surge after listing on the STAR Market. The controversial past of its co-founder, Li Feng, has also resurfaced, including his involvement in the "Mallego Coin" project with Li Xiaolai and others, and a long-standing debt dispute with OKX founder Star involving 1,500 bitcoins (currently worth approximately $135 million). In response, Star recently stated on social media that the debt issue has been handed over to legal action and that the focus should be on the future. 2. Key Data (as of 13:00 HKT, December 8) (Data source: CoinAnk, Upbit, Coingecko, SoSoValue, CoinMarketCap) Bitcoin: $91,596 (down 2.11% year-to-date), daily spot trading volume $40.49 billion. Ethereum: $3,134 (down 6.17% year-to-date), daily spot trading volume $25.27 billion. Fear of Greed Index: 20 (Extreme Fear) Average GAS: BTC: 1.2 sat/vB, ETH: 0.04 Gwei Market share: BTC 58.7%, ETH 12.2% Upbit 24-hour trading volume rankings: XRP, ETH, BTC, MOODENG, SOL 24-hour BTC long/short ratio: 50.54% / 49.46% Sector Performance: Meme and DeFi sectors saw a slight pullback, while SocialFi and AI rose by over 2%. 24-hour liquidation data: A total of 112,699 people worldwide were liquidated, with a total liquidation amount of $416 million. This included $105 million in BTC liquidations, $169 million in ETH liquidations, and $21.92 million in SOL liquidations. 3. ETF Flows (as of December 5) Bitcoin ETFs saw a net outflow of $87.77 million last week, with ARKB experiencing the largest net outflow at $77.86 million. Ethereum ETFs saw net outflows of $65.59 million last week, with BlackRock's ETHA experiencing the largest net outflow at $55.87 million. Solana ETF: Net inflow of $20.3 million last week XRP ETF: Net inflows of $231 million last week, marking the fourth consecutive week of net inflows. 4. Today's Outlook HumidiFi: New token public sale will begin on December 8th at 23:00. The Stable mainnet will launch on December 8th at 21:00. The company formed by the merger of Twenty One Capital and CEP is expected to list on the NYSE on December 9. BounceBit (BB) will unlock approximately 29.93 million tokens at 8:00 AM Beijing time on December 9th, representing 3.42% of the circulating supply, worth approximately $2.7 million. The top 100 cryptocurrencies by market capitalization with the largest gains today are: Ultima up 7%, SPX6900 up 5.8%, Canton Network up 5.5%, Ethena up 5.1%, and Zcash up 4.5%. 5. Hot News Data: APT, LINEA, CHEEL and other tokens will see large-scale unlocking, with APT unlocking value estimated at approximately $19.3 million. This Week's Preview | The Federal Reserve FOMC announces its interest rate decision; the Stable blockchain mainnet will officially launch on December 8th. The largest short position in BTC on Hyperliquid currently has a floating profit of approximately $17 million, having reduced its position by about 20 BTC in 26 minutes. The BEAT team's linked wallet sent $1.2 million worth of tokens to a CEX, seemingly indicating a planned sell-off for profit. Twenty One Capital transferred 43,122 BTC to a new wallet. The U.S. SEC's Cryptocurrency Working Group will hold a roundtable meeting on financial regulation and privacy on December 15. Bittensor will undergo its first halving on December 14th, at which time the daily supply of TAO will decrease to 3600 tokens. ZKsync plans to abandon its early network, ZKsync Lite, in 2026. The long positions held by the "whale that opened short positions after the 1011 flash crash" have reached $164 million, and are currently showing a floating loss of $950,000. A wallet suspected to be Windemute has accumulated approximately $5.2 million worth of SYRUP tokens over the past two weeks. South Korea is considering legislation requiring virtual asset operators to bear "no-fault liability" for hacker attacks, with fines potentially increased to 3% of sales revenue. The average cash cost for public miners mining Bitcoin has reached $74,600, with a total cost of $137,800. Caixin: Last year, 3,032 people were prosecuted for money laundering related to cryptocurrencies; establishing a firewall against virtual currencies is necessary to protect normal economic and trade activities. Farcaster announces strategic shift: from a social-first approach to wallet-driven growth.
Share
PANews2025/12/08 14:48
Robinhood Sets Indonesia Footprint Through Crypto Trader, Brokerage Firms Acquisition
Robinhood Markets has announced two key acquisitions, marking its official entry into the Indonesian market. The American financial services firm has
Share
CryptoNews2025/12/08 14:45