code
stringlengths
17
6.64M
def basic_clean(text): text = ftfy.fix_text(text) text = html.unescape(html.unescape(text)) return text.strip()
def whitespace_clean(text): text = re.sub('\\s+', ' ', text) text = text.strip() return text
class SimpleTokenizer(object): def __init__(self, bpe_path: str=default_bpe()): self.byte_encoder = bytes_to_unicode() self.byte_decoder = {v: k for (k, v) in self.byte_encoder.items()} merges = gzip.open(bpe_path).read().decode('utf-8').split('\n') merges = merges[1:(((49152 - 256) - 2) + 1)] merges = [tuple(merge.split()) for merge in merges] vocab = list(bytes_to_unicode().values()) vocab = (vocab + [(v + '</w>') for v in vocab]) for merge in merges: vocab.append(''.join(merge)) vocab.extend(['<|startoftext|>', '<|endoftext|>']) self.encoder = dict(zip(vocab, range(len(vocab)))) self.decoder = {v: k for (k, v) in self.encoder.items()} self.bpe_ranks = dict(zip(merges, range(len(merges)))) self.cache = {'<|startoftext|>': '<|startoftext|>', '<|endoftext|>': '<|endoftext|>'} self.pat = re.compile("<\\|startoftext\\|>|<\\|endoftext\\|>|'s|'t|'re|'ve|'m|'ll|'d|[\\p{L}]+|[\\p{N}]|[^\\s\\p{L}\\p{N}]+", re.IGNORECASE) def bpe(self, token): if (token in self.cache): return self.cache[token] word = (tuple(token[:(- 1)]) + ((token[(- 1)] + '</w>'),)) pairs = get_pairs(word) if (not pairs): return (token + '</w>') while True: bigram = min(pairs, key=(lambda pair: self.bpe_ranks.get(pair, float('inf')))) if (bigram not in self.bpe_ranks): break (first, second) = bigram new_word = [] i = 0 while (i < len(word)): try: j = word.index(first, i) new_word.extend(word[i:j]) i = j except: new_word.extend(word[i:]) break if ((word[i] == first) and (i < (len(word) - 1)) and (word[(i + 1)] == second)): new_word.append((first + second)) i += 2 else: new_word.append(word[i]) i += 1 new_word = tuple(new_word) word = new_word if (len(word) == 1): break else: pairs = get_pairs(word) word = ' '.join(word) self.cache[token] = word return word def encode(self, text): bpe_tokens = [] text = whitespace_clean(basic_clean(text)).lower() for token in re.findall(self.pat, text): token = ''.join((self.byte_encoder[b] for b in token.encode('utf-8'))) bpe_tokens.extend((self.encoder[bpe_token] for bpe_token in self.bpe(token).split(' '))) return bpe_tokens def decode(self, tokens): text = ''.join([self.decoder[token] for token in tokens]) text = bytearray([self.byte_decoder[c] for c in text]).decode('utf-8', errors='replace').replace('</w>', ' ') return text
@DATASET_REGISTRY.register() class Bamboo(DatasetBase): dataset_dir = 'bamboo' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.image_dir = (root + '/images') self.dataset_dir = root self.preprocessed = os.path.join(self.dataset_dir, 'preprocessed.pkl') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.preprocessed): with open(self.preprocessed, 'rb') as f: preprocessed = pickle.load(f) train = preprocessed['train'] test = preprocessed['test'] else: json_file = (root + '/bamboo_id_map_sample.json') classnames = self.read_classnames(json_file) (train, test, _) = self.read_and_split_data(self.image_dir, p_trn=0.8, ignored=[], new_cnames=classnames) preprocessed = {'train': train, 'test': test} with open(self.preprocessed, 'wb') as f: pickle.dump(preprocessed, f, protocol=pickle.HIGHEST_PROTOCOL) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) train = data['train'] else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) data = {'train': train} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, test) = OxfordPets.subsample_classes(train, test, subsample=subsample) super().__init__(train_x=train, val=test, test=test) @staticmethod def read_classnames(text_file): 'Return a dictionary containing\n key-value pairs of <folder name>: <class name>.\n ' classnames = OrderedDict() import json classnames_origin = json.load(open(text_file, 'r')) for (k, v) in classnames_origin.items(): if isinstance(v, list): classnames[k] = v[0] else: classnames[k] = v return classnames def read_data(self, classnames, split_dir): split_dir = os.path.join(self.image_dir, split_dir) folders = sorted((f.name for f in os.scandir(split_dir) if f.is_dir())) items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(split_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(split_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items @staticmethod def read_and_split_data(image_dir, p_trn=0.5, p_val=0.2, ignored=[], new_cnames=None): categories = listdir_nohidden(image_dir) categories = [c for c in categories if (c not in ignored)] categories.sort() p_tst = ((1 - p_trn) - p_val) print(f'Splitting into {p_trn:.0%} train, {p_val:.0%} val, and {p_tst:.0%} test') def _collate(ims, y, c): items = [] for im in ims: item = Datum(impath=im, label=y, classname=c) items.append(item) return items (train, val, test) = ([], [], []) for (label, category) in enumerate(categories): category_dir = os.path.join(image_dir, category) images = listdir_nohidden(category_dir) images = [os.path.join(category_dir, im) for im in images] random.shuffle(images) n_total = len(images) n_train = round((n_total * p_trn)) n_val = round((n_total * p_val)) n_test = ((n_total - n_train) - n_val) assert (n_train > 0) if ((new_cnames is not None) and (category in new_cnames)): category = new_cnames[category] train.extend(_collate(images[:n_train], label, category)) if (n_val > 0): val.extend(_collate(images[n_train:(n_train + n_val)], label, category)) if (n_test > 0): test.extend(_collate(images[(n_train + n_val):], label, category)) return (train, val, test)
@DATASET_REGISTRY.register() class Caltech101(DatasetBase): dataset_dir = 'caltech-101' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, '101_ObjectCategories') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_Caltech101.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: (train, val, test) = DTD.read_and_split_data(self.image_dir, ignored=IGNORED, new_cnames=NEW_CNAMES) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test)
@DATASET_REGISTRY.register() class DescribableTextures(DatasetBase): dataset_dir = 'dtd' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'images') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_DescribableTextures.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: (train, val, test) = self.read_and_split_data(self.image_dir) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) @staticmethod def read_and_split_data(image_dir, p_trn=0.5, p_val=0.2, ignored=[], new_cnames=None): categories = listdir_nohidden(image_dir) categories = [c for c in categories if (c not in ignored)] categories.sort() p_tst = ((1 - p_trn) - p_val) print(f'Splitting into {p_trn:.0%} train, {p_val:.0%} val, and {p_tst:.0%} test') def _collate(ims, y, c): items = [] for im in ims: item = Datum(impath=im, label=y, classname=c) items.append(item) return items (train, val, test) = ([], [], []) for (label, category) in enumerate(categories): category_dir = os.path.join(image_dir, category) images = listdir_nohidden(category_dir) images = [os.path.join(category_dir, im) for im in images] random.shuffle(images) n_total = len(images) n_train = round((n_total * p_trn)) n_val = round((n_total * p_val)) n_test = ((n_total - n_train) - n_val) assert ((n_train > 0) and (n_val > 0) and (n_test > 0)) if ((new_cnames is not None) and (category in new_cnames)): category = new_cnames[category] train.extend(_collate(images[:n_train], label, category)) val.extend(_collate(images[n_train:(n_train + n_val)], label, category)) test.extend(_collate(images[(n_train + n_val):], label, category)) return (train, val, test)
@DATASET_REGISTRY.register() class EuroSAT(DatasetBase): dataset_dir = 'eurosat' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, '2750') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_EuroSAT.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: (train, val, test) = DTD.read_and_split_data(self.image_dir, new_cnames=NEW_CNAMES) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def update_classname(self, dataset_old): dataset_new = [] for item_old in dataset_old: cname_old = item_old.classname cname_new = NEW_CLASSNAMES[cname_old] item_new = Datum(impath=item_old.impath, label=item_old.label, classname=cname_new) dataset_new.append(item_new) return dataset_new
@DATASET_REGISTRY.register() class FGVCAircraft(DatasetBase): dataset_dir = 'fgvc_aircraft' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'images') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) classnames = [] with open(os.path.join(self.dataset_dir, 'variants.txt'), 'r') as f: lines = f.readlines() for line in lines: classnames.append(line.strip()) cname2lab = {c: i for (i, c) in enumerate(classnames)} train = self.read_data(cname2lab, 'images_variant_train.txt') val = self.read_data(cname2lab, 'images_variant_val.txt') test = self.read_data(cname2lab, 'images_variant_test.txt') num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self, cname2lab, split_file): filepath = os.path.join(self.dataset_dir, split_file) items = [] with open(filepath, 'r') as f: lines = f.readlines() for line in lines: line = line.strip().split(' ') imname = (line[0] + '.jpg') classname = ' '.join(line[1:]) impath = os.path.join(self.image_dir, imname) label = cname2lab[classname] item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class Food101(DatasetBase): dataset_dir = 'food-101' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'images') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_Food101.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: (train, val, test) = DTD.read_and_split_data(self.image_dir) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test)
@DATASET_REGISTRY.register() class ImageNet(DatasetBase): dataset_dir = 'imagenet' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = self.dataset_dir self.preprocessed = os.path.join(self.dataset_dir.replace('group', 'sheng'), 'preprocessed.pkl') self.split_fewshot_dir = os.path.join(self.dataset_dir.replace('group', 'sheng'), 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.preprocessed): with open(self.preprocessed, 'rb') as f: preprocessed = pickle.load(f) train = preprocessed['train'] test = preprocessed['test'] else: text_file = './scripts/classnames.txt' classnames = self.read_classnames(text_file) train = self.read_data(classnames, 'train') test = self.read_data(classnames, 'val') preprocessed = {'train': train, 'test': test} with open(self.preprocessed, 'wb') as f: pickle.dump(preprocessed, f, protocol=pickle.HIGHEST_PROTOCOL) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) train = data['train'] else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) data = {'train': train} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, test) = OxfordPets.subsample_classes(train, test, subsample=subsample) super().__init__(train_x=train, val=test, test=test) @staticmethod def read_classnames(text_file): 'Return a dictionary containing\n key-value pairs of <folder name>: <class name>.\n ' classnames = OrderedDict() with open(text_file, 'r') as f: lines = f.readlines() for line in lines: line = line.strip().split(' ') folder = line[0] classname = ' '.join(line[1:]) classnames[folder] = classname return classnames def read_data(self, classnames, split_dir): split_dir = os.path.join(self.image_dir, split_dir) folders = sorted((f.name for f in os.scandir(split_dir) if f.is_dir())) items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(split_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(split_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class ImageNet21k(DatasetBase): dataset_dir = 'imagenet21k' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.image_dir = root self.dataset_dir = root self.preprocessed = os.path.join(self.dataset_dir.replace('group', 'sheng'), 'preprocessed.pkl') self.split_fewshot_dir = os.path.join(self.dataset_dir.replace('group', 'sheng'), 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.preprocessed): with open(self.preprocessed, 'rb') as f: preprocessed = pickle.load(f) train = preprocessed['train'] test = preprocessed['test'] else: text_file = './scripts/imagenet21k_classnames.txt' classnames = self.read_classnames(text_file) (train, test, _) = self.read_and_split_data(self.image_dir, p_trn=0.8, ignored=[], new_cnames=classnames) preprocessed = {'train': train, 'test': test} with open(self.preprocessed, 'wb') as f: pickle.dump(preprocessed, f, protocol=pickle.HIGHEST_PROTOCOL) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) train = data['train'] else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) data = {'train': train} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, test) = OxfordPets.subsample_classes(train, test, subsample=subsample) super().__init__(train_x=train, val=test, test=test) @staticmethod def read_classnames(text_file): 'Return a dictionary containing\n key-value pairs of <folder name>: <class name>.\n ' classnames = OrderedDict() with open(text_file, 'r') as f: lines = f.readlines() for line in lines: line = line.strip().split(' ') folder = line[0] classname = ' '.join(line[1:]) classnames[folder] = classname return classnames def read_data(self, classnames, split_dir): split_dir = os.path.join(self.image_dir, split_dir) folders = sorted((f.name for f in os.scandir(split_dir) if f.is_dir())) items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(split_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(split_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items @staticmethod def read_and_split_data(image_dir, p_trn=0.5, p_val=0.2, ignored=[], new_cnames=None): categories = listdir_nohidden(image_dir) categories = [c for c in categories if (c not in ignored)] categories.sort() p_tst = ((1 - p_trn) - p_val) print(f'Splitting into {p_trn:.0%} train, {p_val:.0%} val, and {p_tst:.0%} test') def _collate(ims, y, c): items = [] for im in ims: item = Datum(impath=im, label=y, classname=c) items.append(item) return items (train, val, test) = ([], [], []) for (label, category) in enumerate(categories): category_dir = os.path.join(image_dir, category) images = listdir_nohidden(category_dir) images = [os.path.join(category_dir, im) for im in images] random.shuffle(images) n_total = len(images) n_train = round((n_total * p_trn)) n_val = round((n_total * p_val)) n_test = ((n_total - n_train) - n_val) assert (n_train > 0) if ((new_cnames is not None) and (category in new_cnames)): category = new_cnames[category] train.extend(_collate(images[:n_train], label, category)) if (n_val > 0): val.extend(_collate(images[n_train:(n_train + n_val)], label, category)) if (n_test > 0): test.extend(_collate(images[(n_train + n_val):], label, category)) return (train, val, test)
@DATASET_REGISTRY.register() class ImageNetA(DatasetBase): 'ImageNet-A(dversarial).\n\n This dataset is used for testing only.\n ' dataset_dir = 'imagenet-adversarial' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'imagenet-a') text_file = os.path.join(self.dataset_dir, 'classnames.txt') classnames = ImageNet.read_classnames(text_file) data = self.read_data(classnames) super().__init__(train_x=data, test=data) def read_data(self, classnames): image_dir = self.image_dir folders = listdir_nohidden(image_dir, sort=True) folders = [f for f in folders if (f not in TO_BE_IGNORED)] items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(image_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(image_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class ImageNetR(DatasetBase): 'ImageNet-R(endition).\n\n This dataset is used for testing only.\n ' dataset_dir = 'imagenet-rendition' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'imagenet-r') text_file = os.path.join(self.dataset_dir, 'classnames.txt') classnames = ImageNet.read_classnames(text_file) data = self.read_data(classnames) super().__init__(train_x=data, test=data) def read_data(self, classnames): image_dir = self.image_dir folders = listdir_nohidden(image_dir, sort=True) folders = [f for f in folders if (f not in TO_BE_IGNORED)] items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(image_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(image_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class ImageNetSketch(DatasetBase): 'ImageNet-Sketch.\n\n This dataset is used for testing only.\n ' dataset_dir = 'imagenet-sketch' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'images') text_file = os.path.join(self.dataset_dir, 'classnames.txt') classnames = ImageNet.read_classnames(text_file) data = self.read_data(classnames) super().__init__(train_x=data, test=data) def read_data(self, classnames): image_dir = self.image_dir folders = listdir_nohidden(image_dir, sort=True) items = [] for (label, folder) in enumerate(folders): imnames = listdir_nohidden(os.path.join(image_dir, folder)) classname = classnames[folder] for imname in imnames: impath = os.path.join(image_dir, folder, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class ImageNetV2(DatasetBase): 'ImageNetV2.\n\n This dataset is used for testing only.\n ' dataset_dir = 'imagenetv2' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) image_dir = 'imagenetv2-matched-frequency-format-val' self.image_dir = os.path.join(self.dataset_dir, image_dir) text_file = os.path.join(self.dataset_dir, 'classnames.txt') classnames = ImageNet.read_classnames(text_file) data = self.read_data(classnames) super().__init__(train_x=data, test=data) def read_data(self, classnames): image_dir = self.image_dir folders = list(classnames.keys()) items = [] for label in range(1000): class_dir = os.path.join(image_dir, str(label)) imnames = listdir_nohidden(class_dir) folder = folders[label] classname = classnames[folder] for imname in imnames: impath = os.path.join(class_dir, imname) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class OxfordFlowers(DatasetBase): dataset_dir = 'oxford_flowers' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'jpg') self.label_file = os.path.join(self.dataset_dir, 'imagelabels.mat') self.lab2cname_file = os.path.join(self.dataset_dir, 'cat_to_name.json') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_OxfordFlowers.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: (train, val, test) = self.read_data() OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self): tracker = defaultdict(list) label_file = loadmat(self.label_file)['labels'][0] for (i, label) in enumerate(label_file): imname = f'image_{str((i + 1)).zfill(5)}.jpg' impath = os.path.join(self.image_dir, imname) label = int(label) tracker[label].append(impath) print('Splitting data into 50% train, 20% val, and 30% test') def _collate(ims, y, c): items = [] for im in ims: item = Datum(impath=im, label=(y - 1), classname=c) items.append(item) return items lab2cname = read_json(self.lab2cname_file) (train, val, test) = ([], [], []) for (label, impaths) in tracker.items(): random.shuffle(impaths) n_total = len(impaths) n_train = round((n_total * 0.5)) n_val = round((n_total * 0.2)) n_test = ((n_total - n_train) - n_val) assert ((n_train > 0) and (n_val > 0) and (n_test > 0)) cname = lab2cname[str(label)] train.extend(_collate(impaths[:n_train], label, cname)) val.extend(_collate(impaths[n_train:(n_train + n_val)], label, cname)) test.extend(_collate(impaths[(n_train + n_val):], label, cname)) return (train, val, test)
@DATASET_REGISTRY.register() class OxfordPets(DatasetBase): dataset_dir = 'oxford_pets' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'images') self.anno_dir = os.path.join(self.dataset_dir, 'annotations') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_OxfordPets.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = self.read_split(self.split_path, self.image_dir) else: trainval = self.read_data(split_file='trainval.txt') test = self.read_data(split_file='test.txt') (train, val) = self.split_trainval(trainval) self.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = self.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self, split_file): filepath = os.path.join(self.anno_dir, split_file) items = [] with open(filepath, 'r') as f: lines = f.readlines() for line in lines: line = line.strip() (imname, label, species, _) = line.split(' ') breed = imname.split('_')[:(- 1)] breed = '_'.join(breed) breed = breed.lower() imname += '.jpg' impath = os.path.join(self.image_dir, imname) label = (int(label) - 1) item = Datum(impath=impath, label=label, classname=breed) items.append(item) return items @staticmethod def split_trainval(trainval, p_val=0.2): p_trn = (1 - p_val) print(f'Splitting trainval into {p_trn:.0%} train and {p_val:.0%} val') tracker = defaultdict(list) for (idx, item) in enumerate(trainval): label = item.label tracker[label].append(idx) (train, val) = ([], []) for (label, idxs) in tracker.items(): n_val = round((len(idxs) * p_val)) assert (n_val > 0) random.shuffle(idxs) for (n, idx) in enumerate(idxs): item = trainval[idx] if (n < n_val): val.append(item) else: train.append(item) return (train, val) @staticmethod def save_split(train, val, test, filepath, path_prefix): def _extract(items): out = [] for item in items: impath = item.impath label = item.label classname = item.classname impath = impath.replace(path_prefix, '') if impath.startswith('/'): impath = impath[1:] out.append((impath, label, classname)) return out train = _extract(train) val = _extract(val) test = _extract(test) split = {'train': train, 'val': val, 'test': test} write_json(split, filepath) print(f'Saved split to {filepath}') @staticmethod def read_split(filepath, path_prefix): def _convert(items): out = [] for (impath, label, classname) in items: impath = os.path.join(path_prefix, impath) item = Datum(impath=impath, label=int(label), classname=classname) out.append(item) return out print(f'Reading split from {filepath}') split = read_json(filepath) train = _convert(split['train']) val = _convert(split['val']) test = _convert(split['test']) return (train, val, test) @staticmethod def subsample_classes(*args, subsample='all'): 'Divide classes into two groups. The first group\n represents base classes while the second group represents\n new classes.\n\n Args:\n args: a list of datasets, e.g. train, val and test.\n subsample (str): what classes to subsample.\n ' assert (subsample in ['all', 'base', 'new']) if (subsample == 'all'): return args dataset = args[0] labels = set() for item in dataset: labels.add(item.label) labels = list(labels) labels.sort() n = len(labels) m = math.ceil((n / 2)) print(f'SUBSAMPLE {subsample.upper()} CLASSES!') if (subsample == 'base'): selected = labels[:m] else: selected = labels[m:] relabeler = {y: y_new for (y_new, y) in enumerate(selected)} output = [] for dataset in args: dataset_new = [] for item in dataset: if (item.label not in selected): continue item_new = Datum(impath=item.impath, label=relabeler[item.label], classname=item.classname) dataset_new.append(item_new) output.append(dataset_new) return output
@DATASET_REGISTRY.register() class StanfordCars(DatasetBase): dataset_dir = 'stanford_cars' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.split_path = os.path.join(self.dataset_dir, 'split_zhou_StanfordCars.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.dataset_dir) else: trainval_file = os.path.join(self.dataset_dir, 'devkit', 'cars_train_annos.mat') test_file = os.path.join(self.dataset_dir, 'cars_test_annos_withlabels.mat') meta_file = os.path.join(self.dataset_dir, 'devkit', 'cars_meta.mat') trainval = self.read_data('cars_train', trainval_file, meta_file) test = self.read_data('cars_test', test_file, meta_file) (train, val) = OxfordPets.split_trainval(trainval) OxfordPets.save_split(train, val, test, self.split_path, self.dataset_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self, image_dir, anno_file, meta_file): anno_file = loadmat(anno_file)['annotations'][0] meta_file = loadmat(meta_file)['class_names'][0] items = [] for i in range(len(anno_file)): imname = anno_file[i]['fname'][0] impath = os.path.join(self.dataset_dir, image_dir, imname) label = anno_file[i]['class'][(0, 0)] label = (int(label) - 1) classname = meta_file[label][0] names = classname.split(' ') year = names.pop((- 1)) names.insert(0, year) classname = ' '.join(names) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class SUN397(DatasetBase): dataset_dir = 'sun397' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'SUN397') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_SUN397.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: classnames = [] with open(os.path.join(self.dataset_dir, 'ClassName.txt'), 'r') as f: lines = f.readlines() for line in lines: line = line.strip()[1:] classnames.append(line) cname2lab = {c: i for (i, c) in enumerate(classnames)} trainval = self.read_data(cname2lab, 'Training_01.txt') test = self.read_data(cname2lab, 'Testing_01.txt') (train, val) = OxfordPets.split_trainval(trainval) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self, cname2lab, text_file): text_file = os.path.join(self.dataset_dir, text_file) items = [] with open(text_file, 'r') as f: lines = f.readlines() for line in lines: imname = line.strip()[1:] classname = os.path.dirname(imname) label = cname2lab[classname] impath = os.path.join(self.image_dir, imname) names = classname.split('/')[1:] names = names[::(- 1)] classname = ' '.join(names) item = Datum(impath=impath, label=label, classname=classname) items.append(item) return items
@DATASET_REGISTRY.register() class UCF101(DatasetBase): dataset_dir = 'ucf101' def __init__(self, cfg): root = os.path.abspath(os.path.expanduser(cfg.DATASET.ROOT)) self.dataset_dir = os.path.join(root, self.dataset_dir) self.image_dir = os.path.join(self.dataset_dir, 'UCF-101-midframes') self.split_path = os.path.join(self.dataset_dir, 'split_zhou_UCF101.json') self.split_fewshot_dir = os.path.join(self.dataset_dir, 'split_fewshot') mkdir_if_missing(self.split_fewshot_dir) if os.path.exists(self.split_path): (train, val, test) = OxfordPets.read_split(self.split_path, self.image_dir) else: cname2lab = {} filepath = os.path.join(self.dataset_dir, 'ucfTrainTestlist/classInd.txt') with open(filepath, 'r') as f: lines = f.readlines() for line in lines: (label, classname) = line.strip().split(' ') label = (int(label) - 1) cname2lab[classname] = label trainval = self.read_data(cname2lab, 'ucfTrainTestlist/trainlist01.txt') test = self.read_data(cname2lab, 'ucfTrainTestlist/testlist01.txt') (train, val) = OxfordPets.split_trainval(trainval) OxfordPets.save_split(train, val, test, self.split_path, self.image_dir) num_shots = cfg.DATASET.NUM_SHOTS if (num_shots >= 1): seed = cfg.SEED preprocessed = os.path.join(self.split_fewshot_dir, f'shot_{num_shots}-seed_{seed}.pkl') if os.path.exists(preprocessed): print(f'Loading preprocessed few-shot data from {preprocessed}') with open(preprocessed, 'rb') as file: data = pickle.load(file) (train, val) = (data['train'], data['val']) else: train = self.generate_fewshot_dataset(train, num_shots=num_shots) val = self.generate_fewshot_dataset(val, num_shots=min(num_shots, 4)) data = {'train': train, 'val': val} print(f'Saving preprocessed few-shot data to {preprocessed}') with open(preprocessed, 'wb') as file: pickle.dump(data, file, protocol=pickle.HIGHEST_PROTOCOL) subsample = cfg.DATASET.SUBSAMPLE_CLASSES (train, val, test) = OxfordPets.subsample_classes(train, val, test, subsample=subsample) super().__init__(train_x=train, val=val, test=test) def read_data(self, cname2lab, text_file): text_file = os.path.join(self.dataset_dir, text_file) items = [] with open(text_file, 'r') as f: lines = f.readlines() for line in lines: line = line.strip().split(' ')[0] (action, filename) = line.split('/') label = cname2lab[action] elements = re.findall('[A-Z][^A-Z]*', action) renamed_action = '_'.join(elements) filename = filename.replace('.avi', '.jpg') impath = os.path.join(self.image_dir, renamed_action, filename) item = Datum(impath=impath, label=label, classname=renamed_action) items.append(item) return items
def print_args(args, cfg): print('***************') print('** Arguments **') print('***************') optkeys = list(args.__dict__.keys()) optkeys.sort() for key in optkeys: print('{}: {}'.format(key, args.__dict__[key])) print('************') print('** Config **') print('************') print(cfg)
def reset_cfg(cfg, args): if args.root: cfg.DATASET.ROOT = args.root if args.output_dir: cfg.OUTPUT_DIR = args.output_dir if args.trainer: cfg.TRAINER.NAME = args.trainer if args.backbone: cfg.MODEL.BACKBONE.NAME = args.backbone if args.head: cfg.MODEL.HEAD.NAME = args.head
def extend_cfg(cfg): '\n Add new config variables.\n\n E.g.\n from yacs.config import CfgNode as CN\n cfg.TRAINER.MY_MODEL = CN()\n cfg.TRAINER.MY_MODEL.PARAM_A = 1.\n cfg.TRAINER.MY_MODEL.PARAM_B = 0.5\n cfg.TRAINER.MY_MODEL.PARAM_C = False\n ' from yacs.config import CfgNode as CN cfg.TRAINER.OURS = CN() cfg.TRAINER.OURS.N_CTX = 10 cfg.TRAINER.OURS.CSC = False cfg.TRAINER.OURS.CTX_INIT = '' cfg.TRAINER.OURS.WEIGHT_U = 0.1
def setup_cfg(args): cfg = get_cfg_default() extend_cfg(cfg) if args.dataset_config_file: cfg.merge_from_file(args.dataset_config_file) if args.config_file: cfg.merge_from_file(args.config_file) reset_cfg(cfg, args) cfg.freeze() return cfg
def main(args): cfg = setup_cfg(args) if (cfg.SEED >= 0): print('Setting fixed seed: {}'.format(cfg.SEED)) set_random_seed(cfg.SEED) setup_logger(cfg.OUTPUT_DIR) if (torch.cuda.is_available() and cfg.USE_CUDA): torch.backends.cudnn.benchmark = True print_args(args, cfg) print('Collecting env info ...') print('** System info **\n{}\n'.format(collect_env_info())) dataset = eval(cfg.DATASET.NAME)(cfg) if (args.split == 'train'): dataset_input = dataset.train_x elif (args.split == 'val'): dataset_input = dataset.val else: dataset_input = dataset.test tfm_train = build_transform(cfg, is_train=False) data_loader = torch.utils.data.DataLoader(DatasetWrapper(cfg, dataset_input, transform=tfm_train, is_train=False), batch_size=cfg.DATALOADER.TRAIN_X.BATCH_SIZE, sampler=None, shuffle=False, num_workers=cfg.DATALOADER.NUM_WORKERS, drop_last=False, pin_memory=(torch.cuda.is_available() and cfg.USE_CUDA)) (clip_model, _) = clip.load('RN50', 'cuda', jit=False) clip_model.eval() feature_list = [] label_list = [] train_dataiter = iter(data_loader) for train_step in range(1, (len(train_dataiter) + 1)): batch = next(train_dataiter) data = batch['img'].cuda() feature = clip_model.visual(data) feature = feature.cpu() for idx in range(len(data)): feature_list.append(feature[idx].tolist()) label_list.extend(batch['label'].tolist()) save_dir = os.path.join(cfg.OUTPUT_DIR, cfg.DATASET.NAME) os.makedirs(save_dir, exist_ok=True) save_filename = f'{args.split}' np.savez(os.path.join(save_dir, save_filename), feature_list=feature_list, label_list=label_list)
def average_ckpt(state_dict, ignore=['optimizer', 'scheduler']): new_dict = dict() print(state_dict['val_result'], state_dict['epoch']) for key in state_dict: if (key in ignore): continue if isinstance(state_dict[key][0], int): new_dict[key] = int(np.average(state_dict[key])) elif isinstance(state_dict[key][0], float): new_dict[key] = np.average(state_dict[key]) elif isinstance(state_dict[key][0], dict): avg_dict = dict() for ckpt_id in range(len(state_dict[key])): for param_key in state_dict[key][ckpt_id]: if (param_key not in avg_dict): avg_dict[param_key] = [] avg_dict[param_key].append(state_dict[key][ckpt_id][param_key]) for param_key in avg_dict: avg_dict[param_key] = torch.stack(avg_dict[param_key]).mean(dim=0) new_dict[key] = dict(avg_dict) return new_dict
def compute_ci95(res): return ((1.96 * np.std(res)) / np.sqrt(len(res)))
def parse_function(*metrics, directory='', args=None, end_signal=None): print(f'Parsing files in {directory}') subdirs = listdir_nohidden(directory, sort=True) outputs = [] for subdir in subdirs: fpath = osp.join(directory, subdir, 'log.txt') assert check_isfile(fpath) good_to_go = False output = OrderedDict() with open(fpath, 'r') as f: lines = f.readlines() for line in lines: line = line.strip() if (line == end_signal): good_to_go = True for metric in metrics: match = metric['regex'].search(line) if (match and good_to_go): if ('file' not in output): output['file'] = fpath num = float(match.group(1)) name = metric['name'] output[name] = num if output: outputs.append(output) assert (len(outputs) > 0), f'Nothing found in {directory}' metrics_results = defaultdict(list) for output in outputs: msg = '' for (key, value) in output.items(): if isinstance(value, float): msg += f'{key}: {value:.2f}%. ' else: msg += f'{key}: {value}. ' if (key != 'file'): metrics_results[key].append(value) print(msg) output_results = OrderedDict() print('===') print(f'Summary of directory: {directory}') for (key, values) in metrics_results.items(): avg = np.mean(values) std = (compute_ci95(values) if args.ci95 else np.std(values)) print(f'* {key}: {avg:.2f}% +- {std:.2f}%') output_results[key] = avg print('===') return output_results
def main(args, end_signal): metric = {'name': args.keyword, 'regex': re.compile(f'\* {args.keyword}: ([\.\deE+-]+)%')} if args.multi_exp: final_results = defaultdict(list) for directory in listdir_nohidden(args.directory, sort=True): directory = osp.join(args.directory, directory) results = parse_function(metric, directory=directory, args=args, end_signal=end_signal) for (key, value) in results.items(): final_results[key].append(value) print('Average performance') for (key, values) in final_results.items(): avg = np.mean(values) print(f'* {key}: {avg:.2f}%') else: parse_function(metric, directory=args.directory, args=args, end_signal=end_signal)
def main(): with open(f'./scripts/{out_name}.csv', 'w', encoding='UTF8') as f: writer = csv.writer(f) dataset = COOP_ELEVATER_DATASET writer.writerow(([' '] + dataset)) missed = 0 for seed in seeds: temp_row = [] temp_row.append(f'seed {seed}') for data1 in dataset: missed_ = True log_files = glob.glob(f'{ckpt_folder}/{data1}/{ckpt_setting}/seed{seed}/log.txt*') for log_file in log_files: with open(log_file) as open_file: lines = open_file.readlines() number = re.findall('([+-]?[0-9]*\\.[0-9]*)', lines[accuracy_index]) if (('results' in lines[accuracy_index]) and ('test' in lines[(accuracy_index - 2)])): try: temp_row.append(float(number[0])) missed_ = False break except Exception as e: continue if missed_: temp_row.append(' ') missed += 1 print('missed', data1, 'seed', seed) writer.writerow(temp_row) print(f'okay we missed {missed} entries')
def print_args(args, cfg): print('***************') print('** Arguments **') print('***************') optkeys = list(args.__dict__.keys()) optkeys.sort() for key in optkeys: print('{}: {}'.format(key, args.__dict__[key])) print('************') print('** Config **') print('************') print(cfg)
def reset_cfg(cfg, args): if args.root: cfg.DATASET.ROOT = args.root if args.output_dir: cfg.OUTPUT_DIR = args.output_dir if args.resume: cfg.RESUME = args.resume if args.seed: cfg.SEED = args.seed cfg.DATASET.RANDOM_SEED_SAMPLING = args.seed if args.source_domains: cfg.DATASET.SOURCE_DOMAINS = args.source_domains if args.target_domains: cfg.DATASET.TARGET_DOMAINS = args.target_domains if args.transforms: cfg.INPUT.TRANSFORMS = args.transforms if args.trainer: cfg.TRAINER.NAME = args.trainer if args.backbone: cfg.MODEL.BACKBONE.NAME = args.backbone if args.head: cfg.MODEL.HEAD.NAME = args.head if args.dataset: cfg.DATASET.DATASET = args.dataset if args.shots: cfg.DATASET.NUM_SAMPLES_PER_CLASS = args.shots cfg.DATASET.NUM_SHOTS = args.shots if args.multi_task: cfg.DATASET.MULTITASK = args.multi_task if args.multi_task_label_pertask: cfg.DATASET.MULTITASK_LABEL_PERTASK = args.multi_task_label_pertask if args.dataset_coop: cfg.DATASET.COOP = args.dataset_coop if args.cut_contextlen: cfg.TRAINER.CUT_CONTEXTLEN = args.cut_contextlen if args.act_ckpt: cfg.TRAINER.ACT_CKPT = args.act_ckpt if (args.multi_task_evalkey != 'average'): cfg.DATASET.MULTITASK_EVALKEY = args.multi_task_evalkey
def extend_cfg(cfg): '\n Add new config variables.\n\n E.g.\n from yacs.config import CfgNode as CN\n cfg.TRAINER.MY_MODEL = CN()\n cfg.TRAINER.MY_MODEL.PARAM_A = 1.\n cfg.TRAINER.MY_MODEL.PARAM_B = 0.5\n cfg.TRAINER.MY_MODEL.PARAM_C = False\n ' from yacs.config import CfgNode as CN cfg.TRAINER.COOP = CN() cfg.TRAINER.COOP.N_CTX = 16 cfg.TRAINER.COOP.CSC = False cfg.TRAINER.COOP.CTX_INIT = '' cfg.TRAINER.COOP.PREC = 'fp16' cfg.TRAINER.COOP.CLASS_TOKEN_POSITION = 'end' cfg.TRAINER.COCOOP = CN() cfg.TRAINER.COCOOP.N_CTX = 16 cfg.TRAINER.COCOOP.CTX_INIT = '' cfg.TRAINER.COCOOP.PREC = 'fp16' cfg.TRAINER.MVLPT = CN() cfg.TRAINER.MVLPT.PREC = 'fp16' cfg.TRAINER.MVLPT.PROJECT_METHOD = 'transformer' cfg.TRAINER.MVLPT.PROJECT_DIM = 128 cfg.TRAINER.MVLPT.VPT = CN() cfg.TRAINER.MVLPT.VPT.N_CTX = 0 cfg.TRAINER.MVLPT.VPT.CSC = False cfg.TRAINER.MVLPT.VPT.CTX_INIT = '' cfg.TRAINER.MVLPT.VPT.DROPOUT = 0.0 cfg.TRAINER.MVLPT.VPT.PROJECT = (- 1) cfg.TRAINER.MVLPT.VPT.DEEP = True cfg.TRAINER.MVLPT.COOP = CN() cfg.TRAINER.MVLPT.COOP.N_CTX = 0 cfg.TRAINER.MVLPT.COOP.CSC = False cfg.TRAINER.MVLPT.COOP.CTX_INIT = '' cfg.TRAINER.MVLPT.COOP.CLASS_TOKEN_POSITION = 'middle' cfg.TRAINER.MVLPT.COCOOP = CN() cfg.TRAINER.MVLPT.COCOOP.N_CTX = 0 cfg.TRAINER.MVLPT.COCOOP.CTX_INIT = '' cfg.TRAINER.MVLPT.COCOOP.PREC = 'fp16' cfg.DATASET.SUBSAMPLE_CLASSES = 'all' cfg.DATASET.NUM_SAMPLES_PER_CLASS = 20 cfg.DATASET.DATASET = '' cfg.DATASET.RANDOM_SEED_SAMPLING = 1 cfg.DATASET.VAL_SET = '' cfg.DATASET.TRAIN_SET = 'train' cfg.DATASET.TEST_SET = 'val' cfg.DATASET.CENTER_CROP = False cfg.TRAINER.CUT_CONTEXTLEN = False cfg.TRAINER.ACT_CKPT = 1 cfg.DATASET.COOP = False cfg.DATASET.MULTITASK = False cfg.DATASET.MULTITASK_LABEL_PERTASK = False cfg.DATASET.MULTITASK_EVALKEY = 'average'
def setup_cfg(args): cfg = get_cfg_default() extend_cfg(cfg) if args.dataset_config_file: cfg.merge_from_file(args.dataset_config_file) if args.config_file: cfg.merge_from_file(args.config_file) reset_cfg(cfg, args) cfg.merge_from_list(args.opts) cfg.freeze() return cfg
def main(args): cfg = setup_cfg(args) if (cfg.SEED >= 0): print('Setting fixed seed: {}'.format(cfg.SEED)) set_random_seed(cfg.SEED) setup_logger(cfg.OUTPUT_DIR) if (torch.cuda.is_available() and cfg.USE_CUDA): torch.backends.cudnn.benchmark = True print_args(args, cfg) print('Collecting env info ...') print('** System info **\n{}\n'.format(collect_env_info())) trainer = build_trainer(cfg) if args.eval_only: trainer.load_model(args.model_dir, epoch=args.load_epoch) trainer.test() return if args.model_dir: trainer.load_model(args.model_dir) if (not args.no_train): trainer.train()
def add_finetuning_args(parser): parser.add_argument('--ds', required=False, help='Evaluation dataset configure file name.', type=str) parser.add_argument('--model', required=True, help='Evaluation model configure file name', type=str) parser.add_argument('--submit-predictions', help='submit predictions and model info to leaderboard.', default=False, action='store_true') parser.add_argument('--submit-by', help='Person who submits the results.', type=str) parser.add_argument('--no-tuning', help='No hyperparameter-tuning.', default=False, type=(lambda x: (x.lower() == 'true'))) parser.add_argument('--l2', help='(Inverse) L2 regularization strength. This option is only useful when option --no-tuning is True.', default=0.316, type=float) parser.add_argument('--lr', help='Test with a specific learning rate. This option is only useful when option --no-tuning is True.', default=0.001, type=float) parser.add_argument('--run', help='Run id', default=1, type=int) parser.add_argument('--fix_seed', help='Fix the random seed. [-1] not fixing the seeds', default=0, type=int) parser.add_argument('--save-predictions', help='save predictions logits for analysis.', default=True, action='store_true') parser.add_argument('opts', help='Modify config options using the command-line', default=None, nargs=argparse.REMAINDER)
def main(): parser = argparse.ArgumentParser(description='Test a classification model, with finetuning.') add_finetuning_args(parser) args = parser.parse_args() args.cfg = args.ds update_config(config, args) args.cfg = args.model update_config(config, args) config.defrost() config.NAME = '' config.freeze() if args.submit_predictions: assert args.submit_by if (args.fix_seed != (- 1)): random.seed(args.fix_seed) np.random.seed(args.fix_seed) torch.manual_seed(args.fix_seed) torch.cuda.manual_seed_all(args.fix_seed) n_samples = (str(config.DATASET.NUM_SAMPLES_PER_CLASS) if (config.DATASET.NUM_SAMPLES_PER_CLASS > 0) else 'full') exp_name = ('finetuning_' + n_samples) if config.TRAIN.TWO_LR: exp_name += '_two_lr' final_output_dir = create_logger(config, exp_name) if (config.DATASET.NUM_SAMPLES_PER_CLASS == 1): config.defrost() config.DATASET.NUM_SAMPLES_PER_CLASS = 2 config.DATASET.MERGE_TRAIN_VAL_FINAL_RUN = False config.freeze() if comm.is_main_process(): log_arg_env_config(args, config, final_output_dir) if ((config.DATASET.DATASET == 'patch-camelyon') and (config.DATASET.NUM_SAMPLES_PER_CLASS == (- 1))): logging.info(f'Detecting large dataset with {config.DATASET.NUM_SAMPLES_PER_CLASS}-shot.') config.defrost() config.DATASET.NUM_SAMPLES_PER_CLASS = 10000 config.freeze() logging.info(f'Used the subset ({config.DATASET.NUM_SAMPLES_PER_CLASS}-shot) to train the model.') logging.info(f'{config.DATASET.DATASET} is a dataset.') (train_dataloader, val_dataloader, test_dataloader) = construct_dataloader(config) logging.info('Finetuning with full model. This may take several minutes to hours depending on the size of your data.') (best_acc, model_info) = full_model_finetune(train_dataloader, val_dataloader, test_dataloader, args.no_tuning, args.lr, args.l2, config) test_predictions = model_info['best_logits'] if args.save_predictions: import json def json_prec_dump(data, prec=6): return json.dumps(json.loads(json.dumps(data), parse_float=(lambda x: round(float(x), prec)))) results_dict = {'model_name': config.MODEL.NAME, 'dataset_name': config.DATASET.DATASET, 'num_trainable_params': model_info.get('n_trainable_params', None), 'num_params': model_info.get('n_params', None), 'num_visual_params': model_info.get('n_visual_params', None), 'num_backbone_params': model_info.get('n_backbone_params', None), 'n_shot': config.DATASET.NUM_SAMPLES_PER_CLASS, 'rnd_seeds': [config.DATASET.RANDOM_SEED_SAMPLING], 'predictions': [test_predictions.tolist()]} json_string = json_prec_dump(results_dict) prediction_folder = os.path.join(config.OUTPUT_DIR, 'predictions', exp_name) os.makedirs(prediction_folder, exist_ok=True) with open(os.path.join(prediction_folder, f'seed{config.DATASET.RANDOM_SEED_SAMPLING}_{config.DATASET.DATASET}.json'), 'w') as outfile: outfile.write(json_string)
def add_linear_probing_args(parser): parser.add_argument('--ds', required=False, help='Evaluation dataset configure file name.', type=str) parser.add_argument('--model', required=True, help='Evaluation model configure file name', type=str) parser.add_argument('--submit-predictions', help='submit predictions and model info to leaderboard.', default=False, action='store_true') parser.add_argument('--submit-by', help='Person who submits the results.', type=str) parser.add_argument('--no-tuning', help='No hyperparameter-tuning.', default=False, type=(lambda x: (x.lower() == 'true'))) parser.add_argument('--emulate-zeroshot', help='Emulate zero shot learning.', default=False, type=str) parser.add_argument('--l2', help='(Inverse) L2 regularization strength. This option is only useful when option --no-tuning is True.', default=0.316, type=float) parser.add_argument('--lr', help='Test with a specific learning rate. This option is only useful when option --no-tuning is True.', default=0.001, type=float) parser.add_argument('--run', help='Run id', default=1, type=int) parser.add_argument('--fix_seed', help='Fix the random seed. [-1] not fixing the seeds', default=0, type=int) parser.add_argument('--save-predictions', help='save predictions logits for analysis.', default=True, action='store_true') parser.add_argument('opts', help='Modify config options using the command-line', default=None, nargs=argparse.REMAINDER)
def main(): parser = argparse.ArgumentParser(description='Test a classification model, with linear probing.') add_linear_probing_args(parser) args = parser.parse_args() args.cfg = args.ds update_config(config, args) args.cfg = args.model update_config(config, args) config.defrost() config.NAME = '' config.freeze() if args.submit_predictions: assert args.submit_by if (args.fix_seed != (- 1)): random.seed(args.fix_seed) np.random.seed(args.fix_seed) torch.manual_seed(args.fix_seed) torch.cuda.manual_seed_all(args.fix_seed) if args.emulate_zeroshot: args.no_tuning = True config.defrost() config.TRAIN.END_EPOCH = 1 config.TRAIN.EXTRA_FINAL_TRAIN_EPOCH = 0 config.DATASET.NUM_SAMPLES_PER_CLASS = 0 config.TRAIN.EMULATE_ZERO_SHOT = True config.freeze() n_samples = (str(config.DATASET.NUM_SAMPLES_PER_CLASS) if (config.DATASET.NUM_SAMPLES_PER_CLASS >= 0) else 'full') exp_name = ('linear_probe_' + n_samples) if (config.DATASET.NUM_SAMPLES_PER_CLASS == 1): config.defrost() config.DATASET.NUM_SAMPLES_PER_CLASS = 2 config.DATASET.MERGE_TRAIN_VAL_FINAL_RUN = False config.freeze() if config.MODEL.NAME.startswith('mae_'): config.defrost() config.MODEL.SPEC.GLOBAL_POOL = False config.freeze() final_output_dir = create_logger(config, exp_name) if comm.is_main_process(): log_arg_env_config(args, config, final_output_dir) if ((config.DATASET.DATASET == 'patch-camelyon') and (config.DATASET.NUM_SAMPLES_PER_CLASS == (- 1))): logging.info(f'Detecting large dataset with {config.DATASET.NUM_SAMPLES_PER_CLASS}-shot.') config.defrost() config.DATASET.NUM_SAMPLES_PER_CLASS = 10000 config.freeze() logging.info(f'Used the subset ({config.DATASET.NUM_SAMPLES_PER_CLASS}-shot) to train the model.') (train_dataloader, val_dataloader, test_dataloader) = construct_dataloader(config) (best_acc, model_info) = full_model_finetune(train_dataloader, val_dataloader, test_dataloader, args.no_tuning, args.lr, args.l2, config) test_predictions = model_info['best_logits'] if args.save_predictions: import json def json_prec_dump(data, prec=6): return json.dumps(json.loads(json.dumps(data), parse_float=(lambda x: round(float(x), prec)))) results_dict = {'model_name': config.MODEL.NAME, 'dataset_name': config.DATASET.DATASET, 'num_trainable_params': model_info.get('n_trainable_params', None), 'num_params': model_info.get('n_params', None), 'num_visual_params': model_info.get('n_visual_params', None), 'num_backbone_params': model_info.get('n_backbone_params', None), 'n_shot': config.DATASET.NUM_SAMPLES_PER_CLASS, 'rnd_seeds': [config.DATASET.RANDOM_SEED_SAMPLING], 'predictions': [test_predictions.tolist()]} json_string = json_prec_dump(results_dict) prediction_folder = os.path.join(config.OUTPUT_DIR, 'predictions', exp_name) os.makedirs(prediction_folder, exist_ok=True) with open(os.path.join(prediction_folder, f'seed{config.DATASET.RANDOM_SEED_SAMPLING}_{config.DATASET.DATASET}.json'), 'w') as outfile: outfile.write(json_string)
def parse_args(): parser = argparse.ArgumentParser(description='Submit predictions to leaderboard service.') parser.add_argument('--combine_path', required=True, help='Prediction json file path.', type=pathlib.Path) parser.add_argument('--combine_name', default='all_predictions', required=False, help='Output file name.', type=str) args = parser.parse_args() return args
def json_prec_dump(data, prec=6): return json.dumps(json.loads(json.dumps(data), parse_float=(lambda x: round(float(x), prec))))
def main(): logging.basicConfig(level=logging.INFO) args = parse_args() all_predictions = defaultdict(list) for prediction_file in args.combine_path.iterdir(): if (prediction_file.suffix != '.json'): print(f'Ignoring file {prediction_file.name} by suffix.') continue prediction_data = json.loads(prediction_file.read_text()) all_predictions[prediction_data['dataset_name']].append(prediction_data) all_combine_predictions = [] KNOWN_AVERAGE_KEYS = ['num_trainable_params'] KNOWN_MERGE_KEYS = ['rnd_seeds', 'predictions'] KNOWN_DIFF_KEYS = (KNOWN_AVERAGE_KEYS + KNOWN_MERGE_KEYS) for (ds, prediction_data) in all_predictions.items(): prediction_keys = list(prediction_data[0]) combined_dict = dict() for key in prediction_keys: values = [x[key] for x in prediction_data] if (key not in KNOWN_DIFF_KEYS): assert all(((x == values[0]) for x in values)) values = values[0] elif (key in KNOWN_MERGE_KEYS): values = list(itertools.chain.from_iterable(values)) elif (key in KNOWN_AVERAGE_KEYS): values = np.asarray(values).mean() else: assert False combined_dict[key] = values all_combine_predictions.append(combined_dict) all_predictions = {'data': all_combine_predictions} all_predictions = json_prec_dump(all_predictions) save_path = (args.combine_path / f'{args.combine_name}.zip') zf = zipfile.ZipFile(save_path, 'w', zipfile.ZIP_DEFLATED) zf.writestr('all_predictions.json', all_predictions) zf.close()
def add_zero_shot_args(parser): parser.add_argument('--ds', required=False, help='Evaluation dataset configure file name.', type=str) parser.add_argument('--model', required=True, help='Clip model configure file name', type=str) parser.add_argument('--text_feature_only', help='consider text feature or not.', default=False, action='store_true') parser.add_argument('--save-predictions', help='save predictions logits for analysis.', default=True, action='store_true') parser.add_argument('opts', help='Modify config options using the command-line', default=None, nargs=argparse.REMAINDER)
def load_or_extract_features(args, cfg): if (cfg.MODEL.SPEC.TEXT.TOKENIZER == 'clip'): tokenizer = SimpleTokenizer() elif ('hf_' in cfg.MODEL.SPEC.TEXT.TOKENIZER): tokenizer = HFPTTokenizer(pt_name=cfg.MODEL.SPEC.TEXT.TOKENIZER[3:]) else: tokenizer = None feature_file = os.path.join(cfg.DATASET.ROOT, (((('zeroshot_features_' + cfg.MODEL.NAME.replace('/', '')) + f'_wiki_{cfg.KNOWLEDGE.WIKITIONARY.USE_DEFINITION}') + f'_gpt3_{cfg.KNOWLEDGE.GPT3.USE_GPT3}') + '.npy')) logging.info(f'feature_file: {feature_file}') if os.path.exists(feature_file): logging.info('Loading features from existing files.') with open(feature_file, 'rb') as fread: image_features = np.load(fread) text_features = np.load(fread) image_labels = np.load(fread) else: (image_features, image_labels) = extract_features(cfg, test_split_only=True) text_features = extract_text_features(cfg, tokenizer, args) logging.info(f'Test size is {image_features.shape[0]}.') return (image_features, text_features, image_labels)
def load_or_extract_text_features(args, cfg): if (cfg.MODEL.SPEC.TEXT.TOKENIZER == 'clip'): tokenizer = SimpleTokenizer() elif ('hf_' in cfg.MODEL.SPEC.TEXT.TOKENIZER): tokenizer = HFPTTokenizer(pt_name=cfg.MODEL.SPEC.TEXT.TOKENIZER[3:]) else: tokenizer = None feature_file = os.path.join(cfg.DATASET.ROOT, (((('zeroshot_text_features_' + cfg.MODEL.NAME.replace('/', '')) + f'_wiki_{cfg.KNOWLEDGE.WIKITIONARY.USE_DEFINITION}') + f'_gpt3_{cfg.KNOWLEDGE.GPT3.USE_GPT3}') + '.npy')) logging.info(f'feature_file: {feature_file}') if os.path.exists(feature_file): logging.info('Loading features from existing files.') with open(feature_file, 'rb') as fread: text_features = np.load(fread) else: (wiki_dict, gpt3_dict) = extract_text_features(cfg, tokenizer, args) logging.info(f'Test size is {len(wiki_dict)}.') return (wiki_dict, gpt3_dict)
def main(): parser = argparse.ArgumentParser(description='Zero-shot evaluation script.') add_zero_shot_args(parser) args = parser.parse_args() args.cfg = args.ds update_config(config, args) args.cfg = args.model update_config(config, args) config.defrost() config.NAME = '' config.freeze() exp_name = ('zeroshot_eval_' + f'wiki_{config.KNOWLEDGE.WIKITIONARY.USE_DEFINITION}_wnh_{config.KNOWLEDGE.WORDNET.USE_HIERARCHY}_wnd_{config.KNOWLEDGE.WORDNET.USE_DEFINITION}_gpt3_{config.KNOWLEDGE.GPT3.USE_GPT3}') exp_name += f'agg_{config.KNOWLEDGE.AGGREGATION.MEHTOD}_gpt3count_{config.KNOWLEDGE.AGGREGATION.NUM_GPT3_ITEMS}' final_output_dir = create_logger(config, exp_name) if comm.is_main_process(): log_arg_env_config(args, config, final_output_dir) if args.text_feature_only: (wiki_dict, gpt3_dict) = load_or_extract_text_features(args, config) else: (image_features, text_features, image_labels) = load_or_extract_features(args, config) (result, test_predictions, metric) = clip_zeroshot_evaluator(image_features, text_features, image_labels, config) msg = f'=> TEST: {metric} {(100 * result):.3f}% ' logging.info(msg) if args.save_predictions: import json def json_prec_dump(data, prec=6): return json.dumps(json.loads(json.dumps(data), parse_float=(lambda x: round(float(x), prec)))) results_dict = {'model_name': f'CLIP-{config.MODEL.NAME}', 'dataset_name': config.DATASET.DATASET, 'num_trainable_params': 0, 'num_params': config.MODEL.STATS.get('n_params', None), 'num_visual_params': config.MODEL.STATS.get('n_visual_params', None), 'num_backbone_params': config.MODEL.STATS.get('n_backbone_params', None), 'n_shot': 0, 'rnd_seeds': [0], 'predictions': [test_predictions.cpu().data.numpy().tolist()]} json_string = json_prec_dump(results_dict) prediction_folder = os.path.join(config.OUTPUT_DIR, 'predictions', exp_name) os.makedirs(prediction_folder, exist_ok=True) with open(os.path.join(prediction_folder, f'{config.DATASET.DATASET}.json'), 'w') as outfile: outfile.write(json_string)
def get_dataset_hub(): vision_dataset_json = (((pathlib.Path(__file__).resolve().parents[1] / 'resources') / 'datasets') / 'vision_datasets.json').read_text() hub = DatasetHub(vision_dataset_json) return hub
class DataClassBase(): def __post_init__(self): self.validate() @classmethod def from_dict(cls, data_content): c = {} for field in dataclasses.fields(cls): d_type = DataClassBase._get_dataclass_type(field.type) if (field.name in data_content): c[field.name] = (d_type.from_dict(data_content[field.name]) if d_type else data_content[field.name]) assert (len(data_content) == len(c)), f'{data_content.keys()} vs {c.keys()}' return cls(**c) def to_dict(self, skip_default=True): result = {} for f in dataclasses.fields(self): value = getattr(self, f.name) if dataclasses.is_dataclass(value): value = value.to_dict() elif isinstance(value, (list, tuple)): value = type(value)(((v.to_dict() if dataclasses.is_dataclass(v) else v) for v in value)) if ((not skip_default) or (value != f.default)): result[f.name] = value return result def validate(self): for field in dataclasses.fields(self): if (hasattr(field.type, '__origin__') and (field.type.__origin__ in (tuple, collections.abc.Sequence))): expected_types = field.type.__origin__ elif hasattr(field.type, '__args__'): expected_types = field.type.__args__ else: expected_types = field.type if (not isinstance(self.__dict__[field.name], expected_types)): raise TypeError(f'Unexpected field type for {field.name}: Expected: {expected_types}. Actual: {type(self.__dict__[field.name])}') def _raise_value_error(self, config_name, msg=None): error_msg = f'Invalid {config_name}: {getattr(self, config_name)}.' if msg: error_msg += (' ' + msg) raise ValueError(error_msg) def _check_value(self, value_name, checker): value = getattr(self, value_name) if (not checker(value)): raise ValueError(f'Invalid {value_name}: {value}.') def _get_dataclass_type(field_type): 'Returns dataclass type if the given type is dataclass or Optional[dataclass].' if dataclasses.is_dataclass(field_type): return field_type if hasattr(field_type, '__args__'): args = field_type.__args__ if ((len(args) == 2) and (type(None) in args)): return next((t for t in args if dataclasses.is_dataclass(t)), None) return None
class Tasks(): IC_MULTILABEL = DatasetTypes.IC_MULTILABEL IC_MULTICLASS = DatasetTypes.IC_MULTICLASS OBJECT_DETECTION = DatasetTypes.OD VALID_TYPES = [IC_MULTILABEL, IC_MULTICLASS, OBJECT_DETECTION] @staticmethod def is_valid(task): return (task in Tasks.VALID_TYPES)
class Tracks(): LINEAR_PROBING = 'linear_probing' TRANSFER_LEARNING = 'transfer_learning' ZERO_SHOT = 'zero_shot' VALID_TYPES = [LINEAR_PROBING, TRANSFER_LEARNING, ZERO_SHOT] @staticmethod def is_valid(task, track): if (track not in Tracks.VALID_TYPES): return False if (task in [Tasks.IC_MULTICLASS, Tasks.IC_MULTILABEL]): return True if (task == Tasks.OBJECT_DETECTION): return (track != Tracks.LINEAR_PROBING) return False
@dataclasses.dataclass(frozen=True) class PredictionSubmission(DataClassBase): dataset_name: str model_name: str created_by: str task: str track: str predictions: List def validate(self): vision_dataset_json = (((pathlib.Path(__file__).resolve().parents[1] / 'resources') / 'datasets') / 'vision_datasets.json').read_text() hub = DatasetHub(vision_dataset_json) dataset_names = set([x['name'] for x in hub.list_data_version_and_types()]) self._check_value('dataset_name', (lambda x: (x and (x in dataset_names)))) self._check_value('model_name', (lambda x: x)) self._check_value('created_by', (lambda x: x)) self._check_value('task', (lambda x: Tasks.is_valid(x))) self._check_value('track', (lambda x: Tracks.is_valid(self.task, x))) self._check_value('predictions', (lambda x: x)) dataset_manifest = hub.create_dataset_manifest(VISION_DATASET_STORAGE, None, self.dataset_name, usage=Usages.TEST_PURPOSE)[0] logging.info(f'Created test set manifest for {self.dataset_name}') for (fold_idx, predictions) in enumerate(self.predictions): PredictionSubmission.validate_predictions(dataset_manifest, predictions, fold_idx) @staticmethod def validate_predictions(dataset_manifest: DatasetManifest, predictions, fold_idx): assert predictions, f'fold {fold_idx}, empty predictions.' assert (len(predictions) == len(dataset_manifest.images)), f'fold {fold_idx}, Number of predictions does not match number of images.' if (dataset_manifest.data_type in [DatasetTypes.IC_MULTICLASS, DatasetTypes.IC_MULTILABEL]): for (i, probs) in enumerate(predictions): if (dataset_manifest.data_type == DatasetTypes.IC_MULTICLASS): sum_probs = sum(probs) assert math.isclose(sum_probs, 1.0, rel_tol=0.001), f'fold {fold_idx}, Sum of predicted prob vector for image {i}: {sum_probs}, should be 1.0.' assert all([(0.0 <= prob <= 1.0) for prob in probs]), f'fold {fold_idx}, Predicted prob for image {i} not in [0, 1]: {probs}' if (dataset_manifest.data_type == DatasetTypes.OD): for (i, img_wise_bboxes) in enumerate(predictions): for bbox_pred in img_wise_bboxes: assert PredictionSubmission.is_valid_box(bbox_pred, len(dataset_manifest.labelmap)), f'fold {fold_idx}, Invalid predicted bbox for image {i}: {bbox_pred}' @staticmethod def is_valid_box(bbox_pred, num_classes): return ((len(bbox_pred) == 6) and (0 <= bbox_pred[0] < num_classes) and (0.0 <= bbox_pred[1] <= 1.0) and all([(x >= 0) for x in bbox_pred[2:]]) and (bbox_pred[2] <= bbox_pred[4]) and (bbox_pred[3] <= bbox_pred[5]))
@dataclasses.dataclass(frozen=True) class ModelInfoSubmission(DataClassBase): name: str author: str num_params_in_millions: int pretrained_data: str creation_time: str def validate(self): self._check_value('name', (lambda x: x)) self._check_value('author', (lambda x: x)) self._check_value('num_params_in_millions', (lambda x: (x > 0))) self._check_value('pretrained_data', (lambda x: x)) self._check_value('creation_time', (lambda x: datetime.datetime.strptime(x, '%Y-%m-%d')))
def log_arg_env_config(args, config, output_dir): logging.info('=> collecting env info (might take some time)') logging.info(('\n' + get_pretty_env_info())) logging.info(pprint.pformat(args)) logging.info(config) logging.info(f'=> saving logging info into: {output_dir}')
def submit_predictions(prediction_list, submit_by, config, track, task): from vision_benchmark.commands.submit_predictions import submit_predictions_to_leaderboard, submit_model_to_leaderboard submission = {'dataset_name': config.DATASET.DATASET, 'model_name': config.MODEL.NAME, 'track': track, 'task': task, 'created_by': submit_by, 'predictions': [prediction_list]} logging.info('Submit model and predictions to leaderboard.') submit_predictions_to_leaderboard(submission) model_info = {'name': config.MODEL.NAME, 'author': config.MODEL.AUTHOR, 'num_params_in_millions': config.MODEL.NUM_PARAMS_IN_M, 'pretrained_data': config.MODEL.PRETRAINED_DATA, 'creation_time': config.MODEL.CREATION_TIME} submit_model_to_leaderboard(model_info)
def _update_config_from_file(config, cfg_file): config.defrost() with open(cfg_file, 'r') as f: yaml_cfg = yaml.load(f, Loader=yaml.FullLoader) for cfg in yaml_cfg.setdefault('BASE', ['']): if cfg: _update_config_from_file(config, op.join(op.dirname(cfg_file), cfg)) print('=> merge config from {}'.format(cfg_file)) config.merge_from_file(cfg_file) config.freeze()
def update_config(config, args): _update_config_from_file(config, args.cfg) config.defrost() config.merge_from_list(args.opts) config.TRAIN.LR *= comm.world_size (file_name, _) = op.splitext(op.basename(args.cfg)) config.NAME = (file_name + config.NAME) config.RANK = comm.rank if hasattr(config.TRAIN.LR_SCHEDULER, 'METHOD'): if ('timm' == config.TRAIN.LR_SCHEDULER.METHOD): config.TRAIN.LR_SCHEDULER.ARGS.epochs = config.TRAIN.END_EPOCH if ('timm' == config.TRAIN.OPTIMIZER): config.TRAIN.OPTIMIZER_ARGS.lr = config.TRAIN.LR aug = config.AUG if ((aug.MIXUP > 0.0) or (aug.MIXCUT > 0.0) or aug.MIXCUT_MINMAX): aug.MIXUP_PROB = 1.0 config.freeze()
class HFPTTokenizer(object): def __init__(self, pt_name=None): self.pt_name = pt_name self.added_sep_token = 0 self.added_cls_token = 0 self.enable_add_tokens = False self.gpt_special_case = ((not self.enable_add_tokens) and ('gpt' in self.pt_name)) if (pt_name is None): self.tokenizer = AutoTokenizer.from_pretrained('bert-base-cased') else: self.tokenizer = AutoTokenizer.from_pretrained(pt_name) if self.enable_add_tokens: if (self.tokenizer.sep_token is None): self.tokenizer.add_special_tokens({'sep_token': '<SEP>'}) self.added_sep_token = 1 if (self.tokenizer.cls_token is None): self.tokenizer.add_special_tokens({'cls_token': '<CLS>'}) self.added_cls_token = 1 if self.gpt_special_case: self.tokenizer.pad_token = self.tokenizer.eos_token self.tokenizer.sep_token = self.tokenizer.eos_token def get_eot_token(self): return self.tokenizer.encode(self.tokenizer.sep_token, add_special_tokens=False)[0] def get_sot_token(self): return self.tokenizer.encode(self.tokenizer.cls_token, add_special_tokens=False)[0] def get_eot_token_list(self): return self.tokenizer.encode(self.tokenizer.sep_token, add_special_tokens=False) def get_sot_token_list(self): return self.tokenizer.encode(self.tokenizer.cls_token, add_special_tokens=False) def get_tokenizer_obj(self): return self.tokenizer def check_added_tokens(self): return (self.added_sep_token + self.added_cls_token) def tokenize(self, texts: Union[(str, List[str])], context_length: int=77): if isinstance(texts, str): texts = [texts] padding = 'max_length' seqstart = [] seqend = [] max_length = context_length if (self.added_cls_token > 0): seqstart = self.get_sot_token_list() max_length = (max_length - 1) if (self.added_sep_token > 0): seqend = self.get_eot_token_list() max_length = (max_length - 1) tokens = self.tokenizer(texts, padding=padding, truncation=True, max_length=max_length)['input_ids'] for i in range(len(tokens)): tokens[i] = ((seqstart + tokens[i]) + seqend) if self.gpt_special_case: for i in range(len(tokens)): tokens[i][(- 1)] = self.get_eot_token() result = torch.Tensor(tokens).type(torch.LongTensor) return result def get_vocab_size(self): return self.tokenizer.vocab_size def __call__(self, texts: Union[(str, List[str])], context_length: int=77): return self.tokenize(texts, context_length)
def build_tokenizer(tokenizer_name): tokenizer = None if (tokenizer_name == 'clip'): tokenizer = SimpleTokenizer() elif ('hf_' in tokenizer_name): tokenizer = HFPTTokenizer(pt_name=tokenizer_name[3:]) elif ('hfc_' in tokenizer_name): tokenizer = HFPTTokenizer(pt_name=tokenizer_name[4:]) else: raise ValueError('Unknown tokenizer') return tokenizer
class HFPTTokenizer(object): def __init__(self, pt_name=None): self.pt_name = pt_name self.added_sep_token = 0 self.added_cls_token = 0 self.enable_add_tokens = False self.gpt_special_case = ((not self.enable_add_tokens) and ('gpt' in self.pt_name)) if (pt_name is None): self.tokenizer = AutoTokenizer.from_pretrained('bert-base-cased') else: self.tokenizer = AutoTokenizer.from_pretrained(pt_name) if self.enable_add_tokens: if (self.tokenizer.sep_token is None): self.tokenizer.add_special_tokens({'sep_token': '<SEP>'}) self.added_sep_token = 1 if (self.tokenizer.cls_token is None): self.tokenizer.add_special_tokens({'cls_token': '<CLS>'}) self.added_cls_token = 1 if self.gpt_special_case: self.tokenizer.pad_token = self.tokenizer.eos_token self.tokenizer.sep_token = self.tokenizer.eos_token def get_eot_token(self): return self.tokenizer.encode(self.tokenizer.sep_token, add_special_tokens=False)[0] def get_sot_token(self): return self.tokenizer.encode(self.tokenizer.cls_token, add_special_tokens=False)[0] def get_eot_token_list(self): return self.tokenizer.encode(self.tokenizer.sep_token, add_special_tokens=False) def get_sot_token_list(self): return self.tokenizer.encode(self.tokenizer.cls_token, add_special_tokens=False) def get_tokenizer_obj(self): return self.tokenizer def check_added_tokens(self): return (self.added_sep_token + self.added_cls_token) def tokenize(self, texts: Union[(str, List[str])], context_length: int=77): if isinstance(texts, str): texts = [texts] padding = 'max_length' seqstart = [] seqtok = [] seqend = [] max_length = context_length if (self.added_cls_token > 0): seqstart = self.get_sot_token_list() max_length = (max_length - 1) if (self.added_sep_token > 0): seqend = self.get_eot_token_list() max_length = (max_length - 1) tokens = self.tokenizer(texts, padding=padding, truncation=True, max_length=max_length)['input_ids'] for i in range(len(tokens)): tokens[i] = ((seqstart + tokens[i]) + seqend) if self.gpt_special_case: for i in range(len(tokens)): tokens[i][(- 1)] = self.get_eot_token() result = torch.Tensor(tokens).type(torch.LongTensor) return result def get_vocab_size(self): return self.tokenizer.vocab_size def __call__(self, texts: Union[(str, List[str])], context_length: int=77): return self.tokenize(texts, context_length)
def get_prompt_templates(): prompt_templates = ['{}.', 'a photo of a {}.', 'a bad photo of a {}.', 'a photo of many {}.', 'a sculpture of a {}.', 'a photo of the hard to see {}.', 'a low resolution photo of the {}.', 'a rendering of a {}.', 'graffiti of a {}.', 'a bad photo of the {}.', 'a cropped photo of the {}.', 'a tattoo of a {}.', 'the embroidered {}.', 'a photo of a hard to see {}.', 'a bright photo of a {}.', 'a photo of a clean {}.', 'a photo of a dirty {}.', 'a dark photo of the {}.', 'a drawing of a {}.', 'a photo of my {}.', 'the plastic {}.', 'a photo of the cool {}.', 'a close-up photo of a {}.', 'a black and white photo of the {}.', 'a painting of the {}.', 'a painting of a {}.', 'a pixelated photo of the {}.', 'a sculpture of the {}.', 'a bright photo of the {}.', 'a cropped photo of a {}.', 'a plastic {}.', 'a photo of the dirty {}.', 'a jpeg corrupted photo of a {}.', 'a blurry photo of the {}.', 'a photo of the {}.', 'a good photo of the {}.', 'a rendering of the {}.', 'a {} in a video game.', 'a photo of one {}.', 'a doodle of a {}.', 'a close-up photo of the {}.', 'the origami {}.', 'the {} in a video game.', 'a sketch of a {}.', 'a doodle of the {}.', 'a origami {}.', 'a low resolution photo of a {}.', 'the toy {}.', 'a rendition of the {}.', 'a photo of the clean {}.', 'a photo of a large {}.', 'a rendition of a {}.', 'a photo of a nice {}.', 'a photo of a weird {}.', 'a blurry photo of a {}.', 'a cartoon {}.', 'art of a {}.', 'a sketch of the {}.', 'a embroidered {}.', 'a pixelated photo of a {}.', 'itap of the {}.', 'a jpeg corrupted photo of the {}.', 'a good photo of a {}.', 'a plushie {}.', 'a photo of the nice {}.', 'a photo of the small {}.', 'a photo of the weird {}.', 'the cartoon {}.', 'art of the {}.', 'a drawing of the {}.', 'a photo of the large {}.', 'a black and white photo of a {}.', 'the plushie {}.', 'a dark photo of a {}.', 'itap of a {}.', 'graffiti of the {}.', 'a toy {}.', 'itap of my {}.', 'a photo of a cool {}.', 'a photo of a small {}.', 'a tattoo of the {}.'] return prompt_templates
def prompt_engineering(classnames): prompt_templates = get_prompt_templates() temp_idx = np.random.randint(len(prompt_templates)) if isinstance(classnames, list): classname = random.choice(classnames) else: classname = classnames return prompt_templates[temp_idx].replace('{}', classname.replace(',', '').replace('+', ' '))
class Voc2007Classification(torch.utils.data.Dataset): def __init__(self, data_root, image_set='train', transform=None): '\n Pascal voc2007 training/validation data: http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar\n test data: http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar\n ' self.data_root = self._update_path(data_root, image_set) self.transform = transform self.labels = self._read_annotation(image_set) self.images = list(self.labels.keys()) @staticmethod def _update_path(data_root, image_set): if ((image_set == 'train') or (image_set == 'val')): data_root += 'train/VOCdevkit/VOC2007' elif (image_set == 'test'): data_root += 'test/VOCdevkit 2/VOC2007' else: raise Exception('Incorrect image set!') return data_root def __getitem__(self, index): img_path = os.path.join(self.data_root, (('JPEGImages/' + self.images[index]) + '.jpg')) image = Image.open(img_path).convert('RGB') if (self.transform is not None): image = self.transform(image) else: image = transforms.ToTensor()(image) label = self.labels[self.images[index]] label = torch.LongTensor(label) return (image, label) def __len__(self): return len(self.images) def _read_annotation(self, image_set='train'): '\n Annotation interpolation, refer to:\n http://host.robots.ox.ac.uk/pascal/VOC/voc2007/htmldoc/voc.html#SECTION00093000000000000000\n ' object_categories = ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog', 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor'] annotation_folder = os.path.join(self.data_root, 'ImageSets/Main/') files = [file_name for file_name in os.listdir(annotation_folder) if file_name.endswith((('_' + image_set) + '.txt'))] labels_all = dict() for file_name in files: label_str = file_name.split('_')[0] label_int = object_categories.index(label_str) with open(((annotation_folder + '/') + file_name), 'r') as fread: for line in fread.readlines(): index = line[:6] if (index not in labels_all.keys()): labels_all[index] = ([0] * len(object_categories)) flag = 1 if (line[7:9] and (int(line[7:9]) != 1)): flag = (- 1) if (flag == 1): labels_all[index][label_int] = 1 return labels_all
def _is_depthwise(m): return (isinstance(m, nn.Conv2d) and (m.groups == m.in_channels) and (m.groups == m.out_channels))
def _set_wd(cfg, model): without_decay_list = cfg.TRAIN.WITHOUT_WD_LIST without_decay_depthwise = [] without_decay_norm = [] for m in model.modules(): if (_is_depthwise(m) and ('depthwise' in without_decay_list)): without_decay_depthwise.append(m.weight) elif (isinstance(m, nn.BatchNorm2d) and ('bn' in without_decay_list)): without_decay_norm.append(m.weight) without_decay_norm.append(m.bias) elif (isinstance(m, nn.GroupNorm) and ('gn' in without_decay_list)): without_decay_norm.append(m.weight) without_decay_norm.append(m.bias) elif (isinstance(m, nn.LayerNorm) and ('ln' in without_decay_list)): without_decay_norm.append(m.weight) without_decay_norm.append(m.bias) with_decay = [] without_decay = [] skip = {} if hasattr(model, 'no_weight_decay'): skip = model.no_weight_decay() for (n, p) in model.named_parameters(): ever_set = False if (p.requires_grad is False): continue if (n in skip): print('=> set {} wd to 0'.format(n)) without_decay.append(p) continue for pp in without_decay_depthwise: if (p is pp): if cfg.VERBOSE: print('=> set depthwise({}) wd to 0'.format(n)) without_decay.append(p) ever_set = True break for pp in without_decay_norm: if (p is pp): if cfg.VERBOSE: print('=> set norm({}) wd to 0'.format(n)) without_decay.append(p) ever_set = True break if ((not ever_set) and ('bias' in without_decay_list) and n.endswith('.bias')): if cfg.VERBOSE: print('=> set bias({}) wd to 0'.format(n)) without_decay.append(p) elif (not ever_set): with_decay.append(p) params = [{'params': with_decay}, {'params': without_decay, 'weight_decay': 0.0}] return params
def build_optimizer(cfg, model): if (cfg.TRAIN.OPTIMIZER == 'timm'): args = cfg.TRAIN.OPTIMIZER_ARGS print(f'=> usage timm optimizer args: {cfg.TRAIN.OPTIMIZER_ARGS}') optimizer = create_optimizer(args, model) return optimizer optimizer = None params = _set_wd(cfg, model) if (cfg.TRAIN.OPTIMIZER == 'sgd'): if cfg.TRAIN.TWO_LR: trunk_parameters = [] head_parameters = [] for (name, param) in model.named_parameters(): if ('backbone' in name): trunk_parameters.append(param) else: head_parameters.append(param) optimizer = optim.SGD([{'params': trunk_parameters}, {'params': head_parameters, 'lr': cfg.TRAIN.LR}], lr=(cfg.TRAIN.LR * 0.1), momentum=cfg.TRAIN.MOMENTUM, weight_decay=cfg.TRAIN.WD, nesterov=cfg.TRAIN.NESTEROV) else: optimizer = optim.SGD(params, lr=cfg.TRAIN.LR, momentum=cfg.TRAIN.MOMENTUM, weight_decay=cfg.TRAIN.WD, nesterov=cfg.TRAIN.NESTEROV) elif (cfg.TRAIN.OPTIMIZER == 'adam'): if cfg.TRAIN.TWO_LR: trunk_parameters = [] head_parameters = [] for (name, param) in model.named_parameters(): if ('backbone' in name): trunk_parameters.append(param) else: head_parameters.append(param) optimizer = optim.Adam([{'params': trunk_parameters}, {'params': head_parameters, 'lr': cfg.TRAIN.LR}], lr=(cfg.TRAIN.LR * 0.1), weight_decay=cfg.TRAIN.WD) else: optimizer = optim.Adam(params, lr=cfg.TRAIN.LR, weight_decay=cfg.TRAIN.WD) elif (cfg.TRAIN.OPTIMIZER == 'adamW'): optimizer = optim.AdamW(params, lr=cfg.TRAIN.LR, weight_decay=cfg.TRAIN.WD) elif (cfg.TRAIN.OPTIMIZER == 'rmsprop'): optimizer = optim.RMSprop(params, lr=cfg.TRAIN.LR, momentum=cfg.TRAIN.MOMENTUM, weight_decay=cfg.TRAIN.WD, alpha=cfg.TRAIN.RMSPROP_ALPHA, centered=cfg.TRAIN.RMSPROP_CENTERED) return optimizer
class Comm(object): def __init__(self): self.local_rank = 0 @property def world_size(self): if (not dist.is_available()): return 1 if (not dist.is_initialized()): return 1 return dist.get_world_size() @property def rank(self): if (not dist.is_available()): return 0 if (not dist.is_initialized()): return 0 return dist.get_rank() @property def local_rank(self): if (not dist.is_available()): return 0 if (not dist.is_initialized()): return 0 return self._local_rank @local_rank.setter def local_rank(self, value): if (not dist.is_available()): self._local_rank = 0 if (not dist.is_initialized()): self._local_rank = 0 self._local_rank = value @property def head(self): return 'Rank[{}/{}]'.format(self.rank, self.world_size) def is_main_process(self): return (self.rank == 0) def synchronize(self): '\n Helper function to synchronize (barrier) among all processes when\n using distributed training\n ' if (self.world_size == 1): return dist.barrier()
def all_gather(data): '\n Run all_gather on arbitrary picklable data (not necessarily tensors)\n Args:\n data: any picklable object\n Returns:\n list[data]: list of data gathered from each rank\n ' world_size = comm.world_size if (world_size == 1): return [data] buffer = pickle.dumps(data) storage = torch.ByteStorage.from_buffer(buffer) tensor = torch.ByteTensor(storage).to('cuda') local_size = torch.LongTensor([tensor.numel()]).to('cuda') size_list = [torch.LongTensor([0]).to('cuda') for _ in range(world_size)] dist.all_gather(size_list, local_size) size_list = [int(size.item()) for size in size_list] max_size = max(size_list) tensor_list = [] for _ in size_list: tensor_list.append(torch.ByteTensor(size=(max_size,)).to('cuda')) if (local_size != max_size): padding = torch.ByteTensor(size=((max_size - local_size),)).to('cuda') tensor = torch.cat((tensor, padding), dim=0) dist.all_gather(tensor_list, tensor) data_list = [] for (size, tensor) in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list
def reduce_dict(input_dict, average=True): '\n Args:\n input_dict (dict): all the values will be reduced\n average (bool): whether to do average or sum\n Reduce the values in the dictionary from all processes so that process with rank\n 0 has the averaged results. Returns a dict with the same fields as\n input_dict, after reduction.\n ' world_size = comm.world_size if (world_size < 2): return input_dict with torch.no_grad(): names = [] values = [] for k in sorted(input_dict.keys()): names.append(k) values.append(input_dict[k]) values = torch.stack(values, dim=0) dist.reduce(values, dst=0) if ((dist.get_rank() == 0) and average): values /= world_size reduced_dict = {k: v for (k, v) in zip(names, values)} return reduced_dict
def gather_tensors(tensor): '\n Performs all_gather operation on the provided tensors.\n *** Warning ***: torch.distributed.all_gather has no gradient.\n ' tensors_gather = [torch.ones_like(tensor) for _ in range(comm.world_size)] dist.all_gather(tensors_gather, tensor, async_op=False) tensors_gather[comm.rank] = tensor output = torch.cat(tensors_gather, dim=0) return output
def setup_logger(final_output_dir, rank, phase): time_str = time.strftime('%Y-%m-%d-%H-%M') log_file = f'{phase}_{time_str}_rank{rank}.txt' final_log_file = os.path.join(final_output_dir, log_file) head = (('%(asctime)-15s:[P:%(process)d]:' + comm.head) + ' %(message)s') logging.basicConfig(filename=str(final_log_file), format=head) logger = logging.getLogger() logger.setLevel(logging.INFO) console = logging.StreamHandler() console.setFormatter(logging.Formatter(head)) logging.getLogger('').addHandler(console)
def create_logger(cfg, phase='train'): root_output_dir = Path(cfg.OUTPUT_DIR) dataset = cfg.DATASET.DATASET cfg_name = cfg.NAME final_output_dir = ((root_output_dir / dataset) / cfg_name) print('=> creating {} ...'.format(root_output_dir)) root_output_dir.mkdir(parents=True, exist_ok=True) print('=> creating {} ...'.format(final_output_dir)) final_output_dir.mkdir(parents=True, exist_ok=True) print('=> setup logger ...') setup_logger(final_output_dir, cfg.RANK, phase) return str(final_output_dir)
@TRAINER_REGISTRY.register() class ZeroshotCLIP(TrainerX): def build_model(self): cfg = self.cfg classnames = self.dm.dataset.classnames print(f'Loading CLIP (backbone: {cfg.MODEL.BACKBONE.NAME})') clip_model = load_clip_to_cpu(cfg) clip_model.to(self.device) temp = CUSTOM_TEMPLATES[cfg.DATASET.NAME] prompts = [temp.format(c.replace('_', ' ')) for c in classnames] print(f'Prompts: {prompts}') prompts = torch.cat([clip.tokenize(p) for p in prompts]) prompts = prompts.to(self.device) with torch.no_grad(): text_features = clip_model.encode_text(prompts) text_features = (text_features / text_features.norm(dim=(- 1), keepdim=True)) self.text_features = text_features self.clip_model = clip_model def model_inference(self, image): image_features = self.clip_model.encode_image(image) image_features = (image_features / image_features.norm(dim=(- 1), keepdim=True)) logit_scale = self.clip_model.logit_scale.exp() logits = ((logit_scale * image_features) @ self.text_features.t()) return logits
@TRAINER_REGISTRY.register() class ZeroshotCLIP2(ZeroshotCLIP): 'Prompt ensembling.' templates = IMAGENET_TEMPLATES_SELECT def build_model(self): cfg = self.cfg classnames = self.dm.dataset.classnames print(f'Loading CLIP (backbone: {cfg.MODEL.BACKBONE.NAME})') clip_model = load_clip_to_cpu(cfg) clip_model.to(self.device) for params in clip_model.parameters(): params.requires_grad_(False) if (cfg.DATASET.NAME != 'ImageNet'): self.templates += [CUSTOM_TEMPLATES[cfg.DATASET.NAME]] num_temp = len(self.templates) print(f'Prompt ensembling (n={num_temp})') mean_text_features = 0 for (i, temp) in enumerate(self.templates): prompts = [temp.format(c.replace('_', ' ')) for c in classnames] prompts = torch.cat([clip.tokenize(p) for p in prompts]).to(self.device) text_features = clip_model.encode_text(prompts) text_features = (text_features / text_features.norm(dim=(- 1), keepdim=True)) mean_text_features = (mean_text_features + text_features) mean_text_features = (mean_text_features / num_temp) mean_text_features = (mean_text_features / mean_text_features.norm(dim=(- 1), keepdim=True)) self.text_features = mean_text_features self.clip_model = clip_model
def create_config(model_name='u256', timestep_rp=50): if (model_name == 'c64'): config = {'model_path': 'symlink/pretrained/64x64_diffusion.pt', 'classifier_path': 'symlink/pretrained/64x64_classifier.pt', 'image_size': 64, 'batch_size': 64, 'use_ddim': False, 'clip_denoised': True, 'classifier_scale': 0.1} model_config = {'attention_resolutions': '32, 16, 8', 'class_cond': True, 'diffusion_steps': 1000, 'rescale_timesteps': True, 'timestep_respacing': str(timestep_rp), 'dropout': 0.1, 'image_size': 64, 'learn_sigma': True, 'noise_schedule': 'cosine', 'num_channels': 192, 'num_head_channels': 64, 'num_res_blocks': 3, 'resblock_updown': True, 'use_new_attention_order': True, 'use_fp16': True, 'use_scale_shift_norm': True} class_config = {'classifier_depth': 4} elif (model_name == 'c128'): config = {'model_path': 'symlink/pretrained/128x128_diffusion.pt', 'classifier_path': 'symlink/pretrained/128x128_classifier.pt', 'image_size': 128, 'batch_size': 25, 'use_ddim': False, 'clip_denoised': True, 'classifier_scale': 1.25} model_config = {'attention_resolutions': '32, 16, 8', 'class_cond': True, 'diffusion_steps': 1000, 'rescale_timesteps': True, 'timestep_respacing': str(timestep_rp), 'image_size': 128, 'learn_sigma': True, 'noise_schedule': 'linear', 'num_channels': 256, 'num_heads': 4, 'num_res_blocks': 2, 'resblock_updown': True, 'use_fp16': True, 'use_scale_shift_norm': True} class_config = {'image_size': 128} elif (model_name == 'c256'): config = {'model_path': 'symlink/pretrained/256x256_diffusion.pt', 'classifier_path': 'symlink/pretrained/256x256_classifier.pt', 'image_size': 256, 'batch_size': 6, 'use_ddim': False, 'clip_denoised': True, 'classifier_scale': 2.5} model_config = {'attention_resolutions': '32, 16, 8', 'class_cond': True, 'diffusion_steps': 1000, 'rescale_timesteps': True, 'timestep_respacing': str(timestep_rp), 'image_size': 256, 'learn_sigma': True, 'noise_schedule': 'linear', 'num_channels': 256, 'num_head_channels': 64, 'num_res_blocks': 2, 'resblock_updown': True, 'use_fp16': True, 'use_scale_shift_norm': True} class_config = {'image_size': 256} elif (model_name == 'u256'): config = {'model_path': 'symlink/pretrained/256x256_diffusion_uncond.pt', 'classifier_path': 'symlink/pretrained/256x256_classifier.pt', 'image_size': 256, 'batch_size': 20, 'use_ddim': False, 'clip_denoised': True, 'classifier_scale': 10.0} model_config = {'attention_resolutions': '32, 16, 8', 'class_cond': False, 'diffusion_steps': 1000, 'rescale_timesteps': True, 'timestep_respacing': str(timestep_rp), 'image_size': 256, 'learn_sigma': True, 'noise_schedule': 'linear', 'num_channels': 256, 'num_head_channels': 64, 'num_res_blocks': 2, 'resblock_updown': True, 'use_fp16': True, 'use_scale_shift_norm': True} class_config = {'image_size': 256} elif (model_name == 'c512'): config = {'model_path': 'symlink/pretrained/512x512_diffusion.pt', 'classifier_path': 'symlink/pretrained/512x512_classifier.pt', 'image_size': 512, 'batch_size': 9, 'use_ddim': False, 'clip_denoised': True, 'classifier_scale': 9.0} model_config = {'attention_resolutions': '32, 16, 8', 'class_cond': True, 'diffusion_steps': 1000, 'rescale_timesteps': True, 'timestep_respacing': str(timestep_rp), 'image_size': 512, 'learn_sigma': True, 'noise_schedule': 'linear', 'num_channels': 256, 'num_head_channels': 64, 'num_res_blocks': 2, 'resblock_updown': True, 'use_fp16': False, 'use_scale_shift_norm': True} class_config = {'image_size': 512} return (config, model_config, class_config)
def setup_dist(): '\n Setup a distributed process group.\n ' if dist.is_initialized(): return os.environ['CUDA_VISIBLE_DEVICES'] = f'{(MPI.COMM_WORLD.Get_rank() % GPUS_PER_NODE)}' comm = MPI.COMM_WORLD backend = 'gloo' if (backend == 'gloo'): hostname = 'localhost' else: hostname = socket.gethostbyname(socket.getfqdn()) os.environ['MASTER_ADDR'] = comm.bcast(hostname, root=0) os.environ['RANK'] = str(comm.rank) os.environ['WORLD_SIZE'] = str(comm.size) port = comm.bcast(_find_free_port(), root=0) os.environ['MASTER_PORT'] = str(port) dist.init_process_group(backend=backend, init_method='env://')
def dev(): '\n Get the device to use for torch.distributed.\n ' if th.cuda.is_available(): return th.device(f'cuda') return th.device('cpu')
def load_state_dict(path, **kwargs): '\n Load a PyTorch file without redundant fetches across MPI ranks.\n ' chunk_size = (2 ** 30) if (MPI.COMM_WORLD.Get_rank() == 0): with bf.BlobFile(path, 'rb') as f: data = f.read() num_chunks = (len(data) // chunk_size) if (len(data) % chunk_size): num_chunks += 1 MPI.COMM_WORLD.bcast(num_chunks) for i in range(0, len(data), chunk_size): MPI.COMM_WORLD.bcast(data[i:(i + chunk_size)]) else: num_chunks = MPI.COMM_WORLD.bcast(None) data = bytes() for _ in range(num_chunks): data += MPI.COMM_WORLD.bcast(None) return th.load(io.BytesIO(data), **kwargs)
def sync_params(params): '\n Synchronize a sequence of Tensors across ranks from rank 0.\n ' for p in params: with th.no_grad(): dist.broadcast(p, 0)
def _find_free_port(): try: s = socket.socket(socket.AF_INET, socket.SOCK_STREAM) s.bind(('', 0)) s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1) return s.getsockname()[1] finally: s.close()
def convert_module_to_f16(l): '\n Convert primitive modules to float16.\n ' if isinstance(l, (nn.Conv1d, nn.Conv2d, nn.Conv3d)): l.weight.data = l.weight.data.half() if (l.bias is not None): l.bias.data = l.bias.data.half()
def convert_module_to_f32(l): '\n Convert primitive modules to float32, undoing convert_module_to_f16().\n ' if isinstance(l, (nn.Conv1d, nn.Conv2d, nn.Conv3d)): l.weight.data = l.weight.data.float() if (l.bias is not None): l.bias.data = l.bias.data.float()
def make_master_params(param_groups_and_shapes): '\n Copy model parameters into a (differently-shaped) list of full-precision\n parameters.\n ' master_params = [] for (param_group, shape) in param_groups_and_shapes: master_param = nn.Parameter(_flatten_dense_tensors([param.detach().float() for (_, param) in param_group]).view(shape)) master_param.requires_grad = True master_params.append(master_param) return master_params
def model_grads_to_master_grads(param_groups_and_shapes, master_params): '\n Copy the gradients from the model parameters into the master parameters\n from make_master_params().\n ' for (master_param, (param_group, shape)) in zip(master_params, param_groups_and_shapes): master_param.grad = _flatten_dense_tensors([param_grad_or_zeros(param) for (_, param) in param_group]).view(shape)
def master_params_to_model_params(param_groups_and_shapes, master_params): '\n Copy the master parameter data back into the model parameters.\n ' for (master_param, (param_group, _)) in zip(master_params, param_groups_and_shapes): for ((_, param), unflat_master_param) in zip(param_group, unflatten_master_params(param_group, master_param.view((- 1)))): param.detach().copy_(unflat_master_param)
def unflatten_master_params(param_group, master_param): return _unflatten_dense_tensors(master_param, [param for (_, param) in param_group])
def get_param_groups_and_shapes(named_model_params): named_model_params = list(named_model_params) scalar_vector_named_params = ([(n, p) for (n, p) in named_model_params if (p.ndim <= 1)], (- 1)) matrix_named_params = ([(n, p) for (n, p) in named_model_params if (p.ndim > 1)], (1, (- 1))) return [scalar_vector_named_params, matrix_named_params]
def master_params_to_state_dict(model, param_groups_and_shapes, master_params, use_fp16): if use_fp16: state_dict = model.state_dict() for (master_param, (param_group, _)) in zip(master_params, param_groups_and_shapes): for ((name, _), unflat_master_param) in zip(param_group, unflatten_master_params(param_group, master_param.view((- 1)))): assert (name in state_dict) state_dict[name] = unflat_master_param else: state_dict = model.state_dict() for (i, (name, _value)) in enumerate(model.named_parameters()): assert (name in state_dict) state_dict[name] = master_params[i] return state_dict
def state_dict_to_master_params(model, state_dict, use_fp16): if use_fp16: named_model_params = [(name, state_dict[name]) for (name, _) in model.named_parameters()] param_groups_and_shapes = get_param_groups_and_shapes(named_model_params) master_params = make_master_params(param_groups_and_shapes) else: master_params = [state_dict[name] for (name, _) in model.named_parameters()] return master_params
def zero_master_grads(master_params): for param in master_params: param.grad = None
def zero_grad(model_params): for param in model_params: if (param.grad is not None): param.grad.detach_() param.grad.zero_()
def param_grad_or_zeros(param): if (param.grad is not None): return param.grad.data.detach() else: return th.zeros_like(param)
class MixedPrecisionTrainer(): def __init__(self, *, model, use_fp16=False, fp16_scale_growth=0.001, initial_lg_loss_scale=INITIAL_LOG_LOSS_SCALE): self.model = model self.use_fp16 = use_fp16 self.fp16_scale_growth = fp16_scale_growth self.model_params = list(self.model.parameters()) self.master_params = self.model_params self.param_groups_and_shapes = None self.lg_loss_scale = initial_lg_loss_scale if self.use_fp16: self.param_groups_and_shapes = get_param_groups_and_shapes(self.model.named_parameters()) self.master_params = make_master_params(self.param_groups_and_shapes) self.model.convert_to_fp16() def zero_grad(self): zero_grad(self.model_params) def backward(self, loss: th.Tensor): if self.use_fp16: loss_scale = (2 ** self.lg_loss_scale) (loss * loss_scale).backward() else: loss.backward() def optimize(self, opt: th.optim.Optimizer): if self.use_fp16: return self._optimize_fp16(opt) else: return self._optimize_normal(opt) def _optimize_fp16(self, opt: th.optim.Optimizer): logger.logkv_mean('lg_loss_scale', self.lg_loss_scale) model_grads_to_master_grads(self.param_groups_and_shapes, self.master_params) (grad_norm, param_norm) = self._compute_norms(grad_scale=(2 ** self.lg_loss_scale)) if check_overflow(grad_norm): self.lg_loss_scale -= 1 logger.log(f'Found NaN, decreased lg_loss_scale to {self.lg_loss_scale}') zero_master_grads(self.master_params) return False logger.logkv_mean('grad_norm', grad_norm) logger.logkv_mean('param_norm', param_norm) self.master_params[0].grad.mul_((1.0 / (2 ** self.lg_loss_scale))) opt.step() zero_master_grads(self.master_params) master_params_to_model_params(self.param_groups_and_shapes, self.master_params) self.lg_loss_scale += self.fp16_scale_growth return True def _optimize_normal(self, opt: th.optim.Optimizer): (grad_norm, param_norm) = self._compute_norms() logger.logkv_mean('grad_norm', grad_norm) logger.logkv_mean('param_norm', param_norm) opt.step() return True def _compute_norms(self, grad_scale=1.0): grad_norm = 0.0 param_norm = 0.0 for p in self.master_params: with th.no_grad(): param_norm += (th.norm(p, p=2, dtype=th.float32).item() ** 2) if (p.grad is not None): grad_norm += (th.norm(p.grad, p=2, dtype=th.float32).item() ** 2) return ((np.sqrt(grad_norm) / grad_scale), np.sqrt(param_norm)) def master_params_to_state_dict(self, master_params): return master_params_to_state_dict(self.model, self.param_groups_and_shapes, master_params, self.use_fp16) def state_dict_to_master_params(self, state_dict): return state_dict_to_master_params(self.model, state_dict, self.use_fp16)
def check_overflow(value): return ((value == float('inf')) or (value == (- float('inf'))) or (value != value))
def get_named_beta_schedule(schedule_name, num_diffusion_timesteps): '\n Get a pre-defined beta schedule for the given name.\n\n The beta schedule library consists of beta schedules which remain similar\n in the limit of num_diffusion_timesteps.\n Beta schedules may be added, but should not be removed or changed once\n they are committed to maintain backwards compatibility.\n ' if (schedule_name == 'linear'): scale = (1000 / num_diffusion_timesteps) beta_start = (scale * 0.0001) beta_end = (scale * 0.02) return np.linspace(beta_start, beta_end, num_diffusion_timesteps, dtype=np.float64) elif (schedule_name == 'cosine'): return betas_for_alpha_bar(num_diffusion_timesteps, (lambda t: (math.cos(((((t + 0.008) / 1.008) * math.pi) / 2)) ** 2))) else: raise NotImplementedError(f'unknown beta schedule: {schedule_name}')
def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999): '\n Create a beta schedule that discretizes the given alpha_t_bar function,\n which defines the cumulative product of (1-beta) over time from t = [0,1].\n\n :param num_diffusion_timesteps: the number of betas to produce.\n :param alpha_bar: a lambda that takes an argument t from 0 to 1 and\n produces the cumulative product of (1-beta) up to that\n part of the diffusion process.\n :param max_beta: the maximum beta to use; use values lower than 1 to\n prevent singularities.\n ' betas = [] for i in range(num_diffusion_timesteps): t1 = (i / num_diffusion_timesteps) t2 = ((i + 1) / num_diffusion_timesteps) betas.append(min((1 - (alpha_bar(t2) / alpha_bar(t1))), max_beta)) return np.array(betas)
class ModelMeanType(enum.Enum): '\n Which type of output the model predicts.\n ' PREVIOUS_X = enum.auto() START_X = enum.auto() EPSILON = enum.auto()
class ModelVarType(enum.Enum): "\n What is used as the model's output variance.\n\n The LEARNED_RANGE option has been added to allow the model to predict\n values between FIXED_SMALL and FIXED_LARGE, making its job easier.\n " LEARNED = enum.auto() FIXED_SMALL = enum.auto() FIXED_LARGE = enum.auto() LEARNED_RANGE = enum.auto()
class LossType(enum.Enum): MSE = enum.auto() RESCALED_MSE = enum.auto() KL = enum.auto() RESCALED_KL = enum.auto() def is_vb(self): return ((self == LossType.KL) or (self == LossType.RESCALED_KL))
class GaussianDiffusion(): '\n Utilities for training and sampling diffusion models.\n\n Ported directly from here, and then adapted over time to further experimentation.\n https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/diffusion_utils_2.py#L42\n\n :param betas: a 1-D numpy array of betas for each diffusion timestep,\n starting at T and going to 1.\n :param model_mean_type: a ModelMeanType determining what the model outputs.\n :param model_var_type: a ModelVarType determining how variance is output.\n :param loss_type: a LossType determining the loss function to use.\n :param rescale_timesteps: if True, pass floating point timesteps into the\n model so that they are always scaled like in the\n original paper (0 to 1000).\n ' def __init__(self, *, betas, model_mean_type, model_var_type, loss_type, rescale_timesteps=False): self.model_mean_type = model_mean_type self.model_var_type = model_var_type self.loss_type = loss_type self.rescale_timesteps = rescale_timesteps betas = np.array(betas, dtype=np.float64) self.betas = betas assert (len(betas.shape) == 1), 'betas must be 1-D' assert ((betas > 0).all() and (betas <= 1).all()) self.num_timesteps = int(betas.shape[0]) alphas = (1.0 - betas) self.alphas_cumprod = np.cumprod(alphas, axis=0) self.alphas_cumprod_prev = np.append(1.0, self.alphas_cumprod[:(- 1)]) self.alphas_cumprod_next = np.append(self.alphas_cumprod[1:], 0.0) assert (self.alphas_cumprod_prev.shape == (self.num_timesteps,)) self.sqrt_alphas_cumprod = np.sqrt(self.alphas_cumprod) self.sqrt_one_minus_alphas_cumprod = np.sqrt((1.0 - self.alphas_cumprod)) self.log_one_minus_alphas_cumprod = np.log((1.0 - self.alphas_cumprod)) self.sqrt_recip_alphas_cumprod = np.sqrt((1.0 / self.alphas_cumprod)) self.sqrt_recipm1_alphas_cumprod = np.sqrt(((1.0 / self.alphas_cumprod) - 1)) self.posterior_variance = ((betas * (1.0 - self.alphas_cumprod_prev)) / (1.0 - self.alphas_cumprod)) self.posterior_log_variance_clipped = np.log(np.append(self.posterior_variance[1], self.posterior_variance[1:])) self.posterior_mean_coef1 = ((betas * np.sqrt(self.alphas_cumprod_prev)) / (1.0 - self.alphas_cumprod)) self.posterior_mean_coef2 = (((1.0 - self.alphas_cumprod_prev) * np.sqrt(alphas)) / (1.0 - self.alphas_cumprod)) betas = get_named_beta_schedule('linear', 1000) alphas = (1.0 - betas) alphas_cumprod = np.cumprod(alphas, axis=0) g_full = (np.sqrt((1 - alphas_cumprod)) / np.sqrt(alphas_cumprod)) self.g_full = th.tensor(g_full).cuda() self.g = (np.sqrt((1 - self.alphas_cumprod)) / np.sqrt(self.alphas_cumprod)) self.g_prev = (np.sqrt((1 - self.alphas_cumprod_prev)) / np.sqrt(self.alphas_cumprod_prev)) def q_mean_variance(self, x_start, t): "\n Get the distribution q(x_t | x_0).\n\n :param x_start: the [N x C x ...] tensor of noiseless inputs.\n :param t: the number of diffusion steps (minus 1). Here, 0 means one step.\n :return: A tuple (mean, variance, log_variance), all of x_start's shape.\n " mean = (_extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start) variance = _extract_into_tensor((1.0 - self.alphas_cumprod), t, x_start.shape) log_variance = _extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape) return (mean, variance, log_variance) def q_sample(self, x_start, t, noise=None): '\n Diffuse the data for a given number of diffusion steps.\n\n In other words, sample from q(x_t | x_0).\n\n :param x_start: the initial data batch.\n :param t: the number of diffusion steps (minus 1). Here, 0 means one step.\n :param noise: if specified, the split-out normal noise.\n :return: A noisy version of x_start.\n ' if (noise is None): noise = th.randn_like(x_start) assert (noise.shape == x_start.shape) return ((_extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start) + (_extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)) def q_posterior_mean_variance(self, x_start, x_t, t): '\n Compute the mean and variance of the diffusion posterior:\n\n q(x_{t-1} | x_t, x_0)\n\n ' assert (x_start.shape == x_t.shape) posterior_mean = ((_extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start) + (_extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t)) posterior_variance = _extract_into_tensor(self.posterior_variance, t, x_t.shape) posterior_log_variance_clipped = _extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape) assert (posterior_mean.shape[0] == posterior_variance.shape[0] == posterior_log_variance_clipped.shape[0] == x_start.shape[0]) return (posterior_mean, posterior_variance, posterior_log_variance_clipped) def p_mean_variance(self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None): "\n Apply the model to get p(x_{t-1} | x_t), as well as a prediction of\n the initial x, x_0.\n\n :param model: the model, which takes a signal and a batch of timesteps\n as input.\n :param x: the [N x C x ...] tensor at time t.\n :param t: a 1-D Tensor of timesteps.\n :param clip_denoised: if True, clip the denoised signal into [-1, 1].\n :param denoised_fn: if not None, a function which applies to the\n x_start prediction before it is used to sample. Applies before\n clip_denoised.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :return: a dict with the following keys:\n - 'mean': the model mean output.\n - 'variance': the model variance output.\n - 'log_variance': the log of 'variance'.\n - 'pred_xstart': the prediction for x_0.\n " if (model_kwargs is None): model_kwargs = {} (B, C) = x.shape[:2] assert (t.shape == (B,)) model_output = model(x, self._scale_timesteps(t), **model_kwargs) if (self.model_var_type in [ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE]): assert (model_output.shape == (B, (C * 2), *x.shape[2:])) (model_output, model_var_values) = th.split(model_output, C, dim=1) if (self.model_var_type == ModelVarType.LEARNED): model_log_variance = model_var_values model_variance = th.exp(model_log_variance) else: min_log = _extract_into_tensor(self.posterior_log_variance_clipped, t, x.shape) max_log = _extract_into_tensor(np.log(self.betas), t, x.shape) frac = ((model_var_values + 1) / 2) model_log_variance = ((frac * max_log) + ((1 - frac) * min_log)) model_variance = th.exp(model_log_variance) else: (model_variance, model_log_variance) = {ModelVarType.FIXED_LARGE: (np.append(self.posterior_variance[1], self.betas[1:]), np.log(np.append(self.posterior_variance[1], self.betas[1:]))), ModelVarType.FIXED_SMALL: (self.posterior_variance, self.posterior_log_variance_clipped)}[self.model_var_type] model_variance = _extract_into_tensor(model_variance, t, x.shape) model_log_variance = _extract_into_tensor(model_log_variance, t, x.shape) def process_xstart(x): if (denoised_fn is not None): x = denoised_fn(x) if clip_denoised: return x.clamp((- 1), 1) return x if (self.model_mean_type == ModelMeanType.PREVIOUS_X): pred_xstart = process_xstart(self._predict_xstart_from_xprev(x_t=x, t=t, xprev=model_output)) model_mean = model_output elif (self.model_mean_type in [ModelMeanType.START_X, ModelMeanType.EPSILON]): if (self.model_mean_type == ModelMeanType.START_X): pred_xstart = process_xstart(model_output) else: pred_xstart = process_xstart(self._predict_xstart_from_eps(x_t=x, t=t, eps=model_output)) (model_mean, _, _) = self.q_posterior_mean_variance(x_start=pred_xstart, x_t=x, t=t) else: raise NotImplementedError(self.model_mean_type) assert (model_mean.shape == model_log_variance.shape == pred_xstart.shape == x.shape) return {'mean': model_mean, 'variance': model_variance, 'log_variance': model_log_variance, 'pred_xstart': pred_xstart} def p_mean_variance_non_warp(self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None): "\n Apply the model to get p(x_{t-1} | x_t), as well as a prediction of\n the initial x, x_0.\n\n :param model: the model, which takes a signal and a batch of timesteps\n as input.\n :param x: the [N x C x ...] tensor at time t.\n :param t: a 1-D Tensor of timesteps.\n :param clip_denoised: if True, clip the denoised signal into [-1, 1].\n :param denoised_fn: if not None, a function which applies to the\n x_start prediction before it is used to sample. Applies before\n clip_denoised.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :return: a dict with the following keys:\n - 'mean': the model mean output.\n - 'variance': the model variance output.\n - 'log_variance': the log of 'variance'.\n - 'pred_xstart': the prediction for x_0.\n " if (model_kwargs is None): model_kwargs = {} (B, C) = x.shape[:2] assert (t.shape == (B,)) model_output = model(x, self._scale_timesteps(t), **model_kwargs) '\n t = t.long()\n\n if self.model_var_type in [ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE]:\n assert model_output.shape == (B, C * 2, *x.shape[2:])\n model_output, model_var_values = th.split(model_output, C, dim=1)\n if self.model_var_type == ModelVarType.LEARNED:\n model_log_variance = model_var_values\n model_variance = th.exp(model_log_variance)\n else:\n min_log = _extract_into_tensor(\n self.posterior_log_variance_clipped, t, x.shape\n )\n max_log = _extract_into_tensor(np.log(self.betas), t, x.shape)\n # The model_var_values is [-1, 1] for [min_var, max_var].\n frac = (model_var_values + 1) / 2\n model_log_variance = frac * max_log + (1 - frac) * min_log\n model_variance = th.exp(model_log_variance)\n else:\n model_variance, model_log_variance = {\n # for fixedlarge, we set the initial (log-)variance like so\n # to get a better decoder log likelihood.\n ModelVarType.FIXED_LARGE: (\n np.append(self.posterior_variance[1], self.betas[1:]),\n np.log(np.append(self.posterior_variance[1], self.betas[1:])),\n ),\n ModelVarType.FIXED_SMALL: (\n self.posterior_variance,\n self.posterior_log_variance_clipped,\n ),\n }[self.model_var_type]\n model_variance = _extract_into_tensor(model_variance, t, x.shape)\n model_log_variance = _extract_into_tensor(model_log_variance, t, x.shape)\n \n def process_xstart(x):\n if denoised_fn is not None:\n x = denoised_fn(x)\n if clip_denoised:\n return x.clamp(-1, 1)\n return x\n\n if self.model_mean_type == ModelMeanType.PREVIOUS_X:\n pred_xstart = process_xstart(\n self._predict_xstart_from_xprev(x_t=x, t=t, xprev=model_output)\n )\n model_mean = model_output\n elif self.model_mean_type in [ModelMeanType.START_X, ModelMeanType.EPSILON]:\n if self.model_mean_type == ModelMeanType.START_X:\n pred_xstart = process_xstart(model_output)\n else:\n pred_xstart = process_xstart(\n self._predict_xstart_from_eps(x_t=x, t=t, eps=model_output)\n )\n model_mean, _, _ = self.q_posterior_mean_variance(\n x_start=pred_xstart, x_t=x, t=t\n )\n else:\n raise NotImplementedError(self.model_mean_type)\n\n assert (\n model_mean.shape == model_log_variance.shape == pred_xstart.shape == x.shape\n )\n pdb.set_trace()\n ' return {'eps': model_output} def _predict_xstart_from_eps(self, x_t, t, eps): assert (x_t.shape == eps.shape) return ((_extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t) - (_extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * eps)) def _predict_xstart_from_xprev(self, x_t, t, xprev): assert (x_t.shape == xprev.shape) return ((_extract_into_tensor((1.0 / self.posterior_mean_coef1), t, x_t.shape) * xprev) - (_extract_into_tensor((self.posterior_mean_coef2 / self.posterior_mean_coef1), t, x_t.shape) * x_t)) def _predict_eps_from_xstart(self, x_t, t, pred_xstart): return (((_extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t) - pred_xstart) / _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)) def _scale_timesteps(self, t): if self.rescale_timesteps: return (t.float() * (1000.0 / self.num_timesteps)) return t def condition_mean(self, cond_fn, p_mean_var, x, t, model_kwargs=None): '\n Compute the mean for the previous step, given a function cond_fn that\n computes the gradient of a conditional log probability with respect to\n x. In particular, cond_fn computes grad(log(p(y|x))), and we want to\n condition on y.\n\n This uses the conditioning strategy from Sohl-Dickstein et al. (2015).\n ' gradient = cond_fn(x, self._scale_timesteps(t), **model_kwargs) new_mean = (p_mean_var['mean'].float() + (p_mean_var['variance'] * gradient.float())) return new_mean def condition_mean_with_grad(self, cond_fn, p_mean_var, x, t, model_kwargs=None): '\n Compute the mean for the previous step, given a function cond_fn that\n computes the gradient of a conditional log probability with respect to\n x. In particular, cond_fn computes grad(log(p(y|x))), and we want to\n condition on y.\n\n This uses the conditioning strategy from Sohl-Dickstein et al. (2015).\n ' gradient = cond_fn(x, t, p_mean_var, **model_kwargs) new_mean = (p_mean_var['mean'].float() + (p_mean_var['variance'] * gradient.float())) return new_mean def condition_score(self, cond_fn, p_mean_var, x, t, model_kwargs=None): "\n Compute what the p_mean_variance output would have been, should the\n model's score function be conditioned by cond_fn.\n\n See condition_mean() for details on cond_fn.\n\n Unlike condition_mean(), this instead uses the conditioning strategy\n from Song et al (2020).\n " alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape) eps = self._predict_eps_from_xstart(x, t, p_mean_var['pred_xstart']) eps = (eps - ((1 - alpha_bar).sqrt() * cond_fn(x, self._scale_timesteps(t), **model_kwargs))) out = p_mean_var.copy() out['pred_xstart'] = self._predict_xstart_from_eps(x, t, eps) (out['mean'], _, _) = self.q_posterior_mean_variance(x_start=out['pred_xstart'], x_t=x, t=t) return out def condition_score_with_grad(self, cond_fn, p_mean_var, x, t, model_kwargs=None): "\n Compute what the p_mean_variance output would have been, should the\n model's score function be conditioned by cond_fn.\n\n See condition_mean() for details on cond_fn.\n\n Unlike condition_mean(), this instead uses the conditioning strategy\n from Song et al (2020).\n " alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape) eps = self._predict_eps_from_xstart(x, t, p_mean_var['pred_xstart']) eps = (eps - ((1 - alpha_bar).sqrt() * cond_fn(x, t, p_mean_var, **model_kwargs))) out = p_mean_var.copy() out['pred_xstart'] = self._predict_xstart_from_eps(x, t, eps) (out['mean'], _, _) = self.q_posterior_mean_variance(x_start=out['pred_xstart'], x_t=x, t=t) return out def p_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None): "\n Sample x_{t-1} from the model at the given timestep.\n\n :param model: the model to sample from.\n :param x: the current tensor at x_{t-1}.\n :param t: the value of t, starting at 0 for the first diffusion step.\n :param clip_denoised: if True, clip the x_start prediction to [-1, 1].\n :param denoised_fn: if not None, a function which applies to the\n x_start prediction before it is used to sample.\n :param cond_fn: if not None, this is a gradient function that acts\n similarly to the model.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :return: a dict containing the following keys:\n - 'sample': a random sample from the model.\n - 'pred_xstart': a prediction of x_0.\n " out = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) noise = th.randn_like(x) nonzero_mask = (t != 0).float().view((- 1), *([1] * (len(x.shape) - 1))) if (cond_fn is not None): out['mean'] = self.condition_mean(cond_fn, out, x, t, model_kwargs=model_kwargs) sample = (out['mean'] + ((nonzero_mask * th.exp((0.5 * out['log_variance']))) * noise)) return {'sample': sample, 'pred_xstart': out['pred_xstart']} def p_sample_with_grad(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None): "\n Sample x_{t-1} from the model at the given timestep.\n\n :param model: the model to sample from.\n :param x: the current tensor at x_{t-1}.\n :param t: the value of t, starting at 0 for the first diffusion step.\n :param clip_denoised: if True, clip the x_start prediction to [-1, 1].\n :param denoised_fn: if not None, a function which applies to the\n x_start prediction before it is used to sample.\n :param cond_fn: if not None, this is a gradient function that acts\n similarly to the model.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :return: a dict containing the following keys:\n - 'sample': a random sample from the model.\n - 'pred_xstart': a prediction of x_0.\n " with th.enable_grad(): x = x.detach().requires_grad_() out = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) noise = th.randn_like(x) nonzero_mask = (t != 0).float().view((- 1), *([1] * (len(x.shape) - 1))) if (cond_fn is not None): out['mean'] = self.condition_mean_with_grad(cond_fn, out, x, t, model_kwargs=model_kwargs) sample = (out['mean'] + ((nonzero_mask * th.exp((0.5 * out['log_variance']))) * noise)) return {'sample': sample, 'pred_xstart': out['pred_xstart'].detach()} def p_sample_loop(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False): "\n Generate samples from the model.\n\n :param model: the model module.\n :param shape: the shape of the samples, (N, C, H, W).\n :param noise: if specified, the noise from the encoder to sample.\n Should be of the same shape as `shape`.\n :param clip_denoised: if True, clip x_start predictions to [-1, 1].\n :param denoised_fn: if not None, a function which applies to the\n x_start prediction before it is used to sample.\n :param cond_fn: if not None, this is a gradient function that acts\n similarly to the model.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :param device: if specified, the device to create the samples on.\n If not specified, use a model parameter's device.\n :param progress: if True, show a tqdm progress bar.\n :return: a non-differentiable batch of samples.\n " final = None for sample in self.p_sample_loop_progressive(model, shape, noise=noise, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, model_kwargs=model_kwargs, device=device, progress=progress, skip_timesteps=skip_timesteps, init_image=init_image, randomize_class=randomize_class, cond_fn_with_grad=cond_fn_with_grad): final = sample return final['sample'] def p_sample_loop_progressive(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False): '\n Generate samples from the model and yield intermediate samples from\n each timestep of diffusion.\n\n Arguments are the same as p_sample_loop().\n Returns a generator over dicts, where each dict is the return value of\n p_sample().\n ' if (device is None): device = next(model.parameters()).device assert isinstance(shape, (tuple, list)) if (noise is not None): img = noise else: img = th.randn(*shape, device=device) if (skip_timesteps and (init_image is None)): init_image = th.zeros_like(img) indices = list(range((self.num_timesteps - skip_timesteps)))[::(- 1)] if (init_image is not None): my_t = (th.ones([shape[0]], device=device, dtype=th.long) * indices[0]) img = self.q_sample(init_image, my_t, img) if progress: from tqdm.auto import tqdm indices = tqdm(indices) for i in indices: t = th.tensor(([i] * shape[0]), device=device) if (randomize_class and ('y' in model_kwargs)): model_kwargs['y'] = th.randint(low=0, high=model.num_classes, size=model_kwargs['y'].shape, device=model_kwargs['y'].device) with th.no_grad(): sample_fn = (self.p_sample_with_grad if cond_fn_with_grad else self.p_sample) out = sample_fn(model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, model_kwargs=model_kwargs) (yield out) img = out['sample'] def ddim_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None, eta=0.0): '\n Sample x_{t-1} from the model using DDIM.\n\n Same usage as p_sample().\n ' out_orig = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) if (cond_fn is not None): out = self.condition_score(cond_fn, out_orig, x, t, model_kwargs=model_kwargs) else: out = out_orig eps = self._predict_eps_from_xstart(x, t, out['pred_xstart']) alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape) alpha_bar_prev = _extract_into_tensor(self.alphas_cumprod_prev, t, x.shape) sigma = ((eta * th.sqrt(((1 - alpha_bar_prev) / (1 - alpha_bar)))) * th.sqrt((1 - (alpha_bar / alpha_bar_prev)))) noise = th.randn_like(x) mean_pred = ((out['pred_xstart'] * th.sqrt(alpha_bar_prev)) + (th.sqrt(((1 - alpha_bar_prev) - (sigma ** 2))) * eps)) nonzero_mask = (t != 0).float().view((- 1), *([1] * (len(x.shape) - 1))) sample = (mean_pred + ((nonzero_mask * sigma) * noise)) return {'sample': sample, 'pred_xstart': out_orig['pred_xstart']} def ddim_sample_with_grad(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, model_kwargs=None, eta=0.0): '\n Sample x_{t-1} from the model using DDIM.\n\n Same usage as p_sample().\n ' with th.enable_grad(): x = x.detach().requires_grad_() out_orig = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) if (cond_fn is not None): out = self.condition_score_with_grad(cond_fn, out_orig, x, t, model_kwargs=model_kwargs) else: out = out_orig out['pred_xstart'] = out['pred_xstart'].detach() eps = self._predict_eps_from_xstart(x, t, out['pred_xstart']) alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape) alpha_bar_prev = _extract_into_tensor(self.alphas_cumprod_prev, t, x.shape) sigma = ((eta * th.sqrt(((1 - alpha_bar_prev) / (1 - alpha_bar)))) * th.sqrt((1 - (alpha_bar / alpha_bar_prev)))) noise = th.randn_like(x) mean_pred = ((out['pred_xstart'] * th.sqrt(alpha_bar_prev)) + (th.sqrt(((1 - alpha_bar_prev) - (sigma ** 2))) * eps)) nonzero_mask = (t != 0).float().view((- 1), *([1] * (len(x.shape) - 1))) sample = (mean_pred + ((nonzero_mask * sigma) * noise)) return {'sample': sample, 'pred_xstart': out_orig['pred_xstart'].detach()} def ddim_reverse_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None, eta=0.0): '\n Sample x_{t+1} from the model using DDIM reverse ODE.\n ' assert (eta == 0.0), 'Reverse ODE only for deterministic path' out = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) eps = (((_extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x.shape) * x) - out['pred_xstart']) / _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x.shape)) alpha_bar_next = _extract_into_tensor(self.alphas_cumprod_next, t, x.shape) mean_pred = ((out['pred_xstart'] * th.sqrt(alpha_bar_next)) + (th.sqrt((1 - alpha_bar_next)) * eps)) return {'sample': mean_pred, 'pred_xstart': out['pred_xstart']} def ddim_sample_loop(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, eta=0.0, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False): '\n Generate samples from the model using DDIM.\n\n Same usage as p_sample_loop().\n ' final = None for sample in self.ddim_sample_loop_progressive(model, shape, noise=noise, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, model_kwargs=model_kwargs, device=device, progress=progress, eta=eta, skip_timesteps=skip_timesteps, init_image=init_image, randomize_class=randomize_class, cond_fn_with_grad=cond_fn_with_grad): final = sample return final['sample'] def ddim_sample_loop_progressive(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, eta=0.0, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False): '\n Use DDIM to sample from the model and yield intermediate samples from\n each timestep of DDIM.\n\n Same usage as p_sample_loop_progressive().\n ' if (device is None): device = next(model.parameters()).device assert isinstance(shape, (tuple, list)) if (noise is not None): img = noise else: img = th.randn(*shape, device=device) if (skip_timesteps and (init_image is None)): init_image = th.zeros_like(img) indices = list(range((self.num_timesteps - skip_timesteps)))[::(- 1)] if (init_image is not None): my_t = (th.ones([shape[0]], device=device, dtype=th.long) * indices[0]) img = self.q_sample(init_image, my_t, img) if progress: from fastprogress import progress_bar if isinstance(progress, bool): indices = progress_bar(indices) else: indices = progress_bar(indices, parent=progress) for i in indices: t = th.tensor(([i] * shape[0]), device=device) if (randomize_class and ('y' in model_kwargs)): model_kwargs['y'] = th.randint(low=0, high=model.num_classes, size=model_kwargs['y'].shape, device=model_kwargs['y'].device) with th.no_grad(): sample_fn = (self.ddim_sample_with_grad if cond_fn_with_grad else self.ddim_sample) out = sample_fn(model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, model_kwargs=model_kwargs, eta=eta) (yield out) img = out['sample'] def plms_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, cond_fn_with_grad=False, order=2, old_out=None): '\n Sample x_{t-1} from the model using Pseudo Linear Multistep.\n\n Same usage as p_sample().\n ' if ((not int(order)) or (not (1 <= order <= 4))): raise ValueError('order is invalid (should be int from 1-4).') def get_model_output(x, t): with th.set_grad_enabled((cond_fn_with_grad and (cond_fn is not None))): x = (x.detach().requires_grad_() if cond_fn_with_grad else x) out_orig = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) if (cond_fn is not None): if cond_fn_with_grad: out = self.condition_score_with_grad(cond_fn, out_orig, x, t, model_kwargs=model_kwargs) x = x.detach() else: out = self.condition_score(cond_fn, out_orig, x, t, model_kwargs=model_kwargs) else: out = out_orig eps = self._predict_eps_from_xstart(x, t, out['pred_xstart']) return (eps, out, out_orig) alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape) alpha_bar_prev = _extract_into_tensor(self.alphas_cumprod_prev, t, x.shape) (eps, out, out_orig) = get_model_output(x, t) if ((order >= 1) and (old_out is None)): old_eps = [eps] mean_pred = ((out['pred_xstart'] * th.sqrt(alpha_bar_prev)) + (th.sqrt((1 - alpha_bar_prev)) * eps)) (eps_2, _, _) = get_model_output(mean_pred, (t - 1)) eps_prime = ((eps + eps_2) / 2) pred_prime = self._predict_xstart_from_eps(x, t, eps_prime) mean_pred = ((pred_prime * th.sqrt(alpha_bar_prev)) + (th.sqrt((1 - alpha_bar_prev)) * eps_prime)) else: old_eps = old_out['old_eps'] old_eps.append(eps) cur_order = min(order, len(old_eps)) if (cur_order == 1): eps_prime = old_eps[(- 1)] elif (cur_order == 2): eps_prime = (((3 * old_eps[(- 1)]) - old_eps[(- 2)]) / 2) elif (cur_order == 3): eps_prime = ((((23 * old_eps[(- 1)]) - (16 * old_eps[(- 2)])) + (5 * old_eps[(- 3)])) / 12) elif (cur_order == 4): eps_prime = (((((55 * old_eps[(- 1)]) - (59 * old_eps[(- 2)])) + (37 * old_eps[(- 3)])) - (9 * old_eps[(- 4)])) / 24) else: raise RuntimeError('cur_order is invalid.') pred_prime = self._predict_xstart_from_eps(x, t, eps_prime) mean_pred = ((pred_prime * th.sqrt(alpha_bar_prev)) + (th.sqrt((1 - alpha_bar_prev)) * eps_prime)) if (len(old_eps) >= order): old_eps.pop(0) nonzero_mask = (t != 0).float().view((- 1), *([1] * (len(x.shape) - 1))) sample = ((mean_pred * nonzero_mask) + (out['pred_xstart'] * (1 - nonzero_mask))) if (impu_fn is not None): sample = self.condition_score3(impu_fn, None, sample, t, model_kwargs=model_kwargs) return {'sample': sample, 'pred_xstart': out_orig['pred_xstart'], 'old_eps': old_eps} def plms_sample_loop(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Generate samples from the model using Pseudo Linear Multistep.\n\n Same usage as p_sample_loop().\n ' final = None for sample in self.plms_sample_loop_progressive(model, shape, noise=noise, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=inpu_fn, model_kwargs=model_kwargs, device=device, progress=progress, skip_timesteps=skip_timesteps, init_image=init_image, randomize_class=randomize_class, cond_fn_with_grad=cond_fn_with_grad, order=order): final = sample return final['sample'] def plms_sample_loop_progressive(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Use PLMS to sample from the model and yield intermediate samples from each\n timestep of PLMS.\n\n Same usage as p_sample_loop_progressive().\n ' if (device is None): device = next(model.parameters()).device assert isinstance(shape, (tuple, list)) if (noise is not None): img = noise else: img = th.randn(*shape, device=device) if (skip_timesteps and (init_image is None)): init_image = th.zeros_like(img) indices = list(range((self.num_timesteps - skip_timesteps)))[::(- 1)] if (init_image is not None): my_t = (th.ones([shape[0]], device=device, dtype=th.long) * indices[0]) img = self.q_sample(init_image, my_t, img) if progress: from fastprogress import progress_bar if isinstance(progress, bool): indices = progress_bar(indices) else: indices = progress_bar(indices, parent=progress) old_out = None for i in indices: t = th.tensor(([i] * shape[0]), device=device) if (randomize_class and ('y' in model_kwargs)): model_kwargs['y'] = th.randint(low=0, high=model.num_classes, size=model_kwargs['y'].shape, device=model_kwargs['y'].device) with th.no_grad(): out = self.plms_sample(model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=impu_fn, model_kwargs=model_kwargs, cond_fn_with_grad=cond_fn_with_grad, order=order, old_out=old_out) (yield out) old_out = out img = out['sample'] def condition_score2(self, cond_fn, p_mean_var, x, t, del_g, s0=0, s_1=0, model_kwargs=None): '\n Unlike condition_score(), this function output only grad value.\n Note that p_mean_var and s_1 is never used in this version.\n ' grad = cond_fn(x, self._scale_timesteps(t), **model_kwargs) return (((- grad) * del_g) * (1 - (s0 ** 2)).sqrt()) def condition_score3(self, cond_fn, p_mean_var, x, t, model_kwargs=None): x = cond_fn(x, self._scale_timesteps(t), **model_kwargs) return x def stsp_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, cond_fn_with_grad=False, order=2, old_out=None): '\n Sample x_{t-1} from the model using Pseudo Linear Multistep\n and Strange Splitting for conditioning.\n Same usage as p_sample().\n ' g0 = _extract_into_tensor(self.g, t[0], (1,)) g_1 = _extract_into_tensor(self.g_prev, t[0], (1,)) s0 = (1 / th.sqrt(((g0 ** 2) + 1))) s_1 = (1 / th.sqrt(((g_1 ** 2) + 1))) del_g = (g_1 - g0) if (cond_fn is not None): alpha_half = (1 / ((((g0 + g_1) ** 2) / 4) + 1)) s_half = th.sqrt(alpha_half) grad = self.condition_score2(cond_fn, None, x, t, (del_g / 2), s0, s_half, model_kwargs=model_kwargs) x = (x + (grad * s0)) out_orig = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) eps = self._predict_eps_from_xstart(x, t, out_orig['pred_xstart']) if (old_out is None): old_out = [] old_eps = [eps] else: old_eps = old_out['old_eps'] old_eps.append(eps) eps_prime = plms_mixer(old_eps, order) sample = (((x / s0) + (del_g * eps_prime)) * s_1) if (cond_fn is not None): if (t[0].long() < 1): t = (t + 1) grad = self.condition_score2(cond_fn, out_orig, sample, (t - 1), (del_g / 2), s_half, s_1, model_kwargs=model_kwargs) sample = (sample + (grad * s_1)) if (impu_fn is not None): sample = self.condition_score3(impu_fn, None, sample, t, model_kwargs=model_kwargs) return {'sample': sample, 'pred_xstart': out_orig['pred_xstart'], 'old_eps': old_eps} def stsp_sample_loop(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Generate samples from the model using Strang Splitting \n and Pseudo Linear Multistep.\n\n Same usage as p_sample_loop().\n ' final = None for sample in self.stsp_sample_loop_progressive(model, shape, noise=noise, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=impu_fn, model_kwargs=model_kwargs, device=device, progress=progress, skip_timesteps=skip_timesteps, init_image=init_image, randomize_class=randomize_class, cond_fn_with_grad=cond_fn_with_grad, order=order): final = sample return final['sample'] def stsp_sample_loop_progressive(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Use STSP to sample from the model and yield intermediate samples from each\n timestep of STSP.\n\n Same usage as p_sample_loop_progressive().\n ' if (device is None): device = next(model.parameters()).device assert isinstance(shape, (tuple, list)) if (noise is not None): img = noise else: img = th.randn(*shape, device=device) if (skip_timesteps and (init_image is None)): init_image = th.zeros_like(img) indices = list(range((self.num_timesteps - skip_timesteps)))[::(- 1)] if (init_image is not None): my_t = (th.ones([shape[0]], device=device, dtype=th.long) * indices[0]) img = self.q_sample(init_image, my_t, img) if progress: from fastprogress import progress_bar if isinstance(progress, bool): indices = progress_bar(indices) else: indices = progress_bar(indices, parent=progress) old_out = None for i in indices: t = th.tensor(([i] * shape[0]), device=device) if (randomize_class and ('y' in model_kwargs)): model_kwargs['y'] = th.randint(low=0, high=model.num_classes, size=model_kwargs['y'].shape, device=model_kwargs['y'].device) with th.no_grad(): out = self.stsp_sample(model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=impu_fn, model_kwargs=model_kwargs, order=order, old_out=old_out) (yield out) old_out = out img = out['sample'] def ltsp_sample(self, model, x, t, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, cond_fn_with_grad=False, order=2, old_out=None): '\n Sample x_{t-1} from the model using Lie-Trotter Splitting \n and Pseudo Linear Multistep.\n\n Same usage as p_sample().\n ' g0 = _extract_into_tensor(self.g, t[0], (1,)) g_1 = _extract_into_tensor(self.g_prev, t[0], (1,)) s0 = (1 / th.sqrt(((g0 ** 2) + 1))) s_1 = (1 / th.sqrt(((g_1 ** 2) + 1))) del_g = (g_1 - g0) if ((cond_fn is not None) and True): grad = self.condition_score2(cond_fn, None, x, t, del_g, s0, s_1, model_kwargs=model_kwargs) x = (x + (grad * s0)) out_orig = self.p_mean_variance(model, x, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs) eps = self._predict_eps_from_xstart(x, t, out_orig['pred_xstart']) if (old_out is None): old_out = [] old_eps = [eps] else: old_eps = old_out['old_eps'] old_eps.append(eps) eps_prime = plms_mixer(old_eps, order) sample = (((x / s0) + (del_g * eps_prime)) * s_1) if ((cond_fn is not None) and False): grad = self.condition_score2(cond_fn, None, sample, t, del_g, s0, s_1, model_kwargs=model_kwargs) sample = (sample + (grad * s_1)) if (impu_fn is not None): sample = self.condition_score3(impu_fn, None, sample, t, model_kwargs=model_kwargs) return {'sample': sample, 'pred_xstart': out_orig['pred_xstart'], 'old_eps': old_eps} def ltsp_sample_loop(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Generate samples from the model using Lie-Trotter Splitting \n and Pseudo Linear Multistep.\n\n Same usage as p_sample_loop().\n ' final = None for sample in self.ltsp_sample_loop_progressive(model, shape, noise=noise, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=impu_fn, model_kwargs=model_kwargs, device=device, progress=progress, skip_timesteps=skip_timesteps, init_image=init_image, randomize_class=randomize_class, cond_fn_with_grad=cond_fn_with_grad, order=order): final = sample return final['sample'] def ltsp_sample_loop_progressive(self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, cond_fn=None, impu_fn=None, model_kwargs=None, device=None, progress=False, skip_timesteps=0, init_image=None, randomize_class=False, cond_fn_with_grad=False, order=2): '\n Use LTPS to sample from the model and yield intermediate samples from each\n timestep of LTSP.\n\n Same usage as p_sample_loop_progressive().\n ' if (device is None): device = next(model.parameters()).device assert isinstance(shape, (tuple, list)) if (noise is not None): img = noise else: img = th.randn(*shape, device=device) if (skip_timesteps and (init_image is None)): init_image = th.zeros_like(img) indices = list(range((self.num_timesteps - skip_timesteps)))[::(- 1)] if (init_image is not None): my_t = (th.ones([shape[0]], device=device, dtype=th.long) * indices[0]) img = self.q_sample(init_image, my_t, img) if progress: from fastprogress import progress_bar if isinstance(progress, bool): indices = progress_bar(indices) else: indices = progress_bar(indices, parent=progress) old_out = None for i in indices: t = th.tensor(([i] * shape[0]), device=device) if (randomize_class and ('y' in model_kwargs)): model_kwargs['y'] = th.randint(low=0, high=model.num_classes, size=model_kwargs['y'].shape, device=model_kwargs['y'].device) with th.no_grad(): out = self.ltsp_sample(model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, cond_fn=cond_fn, impu_fn=impu_fn, model_kwargs=model_kwargs, order=order, old_out=old_out) (yield out) old_out = out img = out['sample'] def _vb_terms_bpd(self, model, x_start, x_t, t, clip_denoised=True, model_kwargs=None): "\n Get a term for the variational lower-bound.\n\n The resulting units are bits (rather than nats, as one might expect).\n This allows for comparison to other papers.\n\n :return: a dict with the following keys:\n - 'output': a shape [N] tensor of NLLs or KLs.\n - 'pred_xstart': the x_0 predictions.\n " (true_mean, _, true_log_variance_clipped) = self.q_posterior_mean_variance(x_start=x_start, x_t=x_t, t=t) out = self.p_mean_variance(model, x_t, t, clip_denoised=clip_denoised, model_kwargs=model_kwargs) kl = normal_kl(true_mean, true_log_variance_clipped, out['mean'], out['log_variance']) kl = (mean_flat(kl) / np.log(2.0)) decoder_nll = (- discretized_gaussian_log_likelihood(x_start, means=out['mean'], log_scales=(0.5 * out['log_variance']))) assert (decoder_nll.shape == x_start.shape) decoder_nll = (mean_flat(decoder_nll) / np.log(2.0)) output = th.where((t == 0), decoder_nll, kl) return {'output': output, 'pred_xstart': out['pred_xstart']} def training_losses(self, model, x_start, t, model_kwargs=None, noise=None): '\n Compute training losses for a single timestep.\n\n :param model: the model to evaluate loss on.\n :param x_start: the [N x C x ...] tensor of inputs.\n :param t: a batch of timestep indices.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n :param noise: if specified, the specific Gaussian noise to try to remove.\n :return: a dict with the key "loss" containing a tensor of shape [N].\n Some mean or variance settings may also have other keys.\n ' if (model_kwargs is None): model_kwargs = {} if (noise is None): noise = th.randn_like(x_start) x_t = self.q_sample(x_start, t, noise=noise) terms = {} if ((self.loss_type == LossType.KL) or (self.loss_type == LossType.RESCALED_KL)): terms['loss'] = self._vb_terms_bpd(model=model, x_start=x_start, x_t=x_t, t=t, clip_denoised=False, model_kwargs=model_kwargs)['output'] if (self.loss_type == LossType.RESCALED_KL): terms['loss'] *= self.num_timesteps elif ((self.loss_type == LossType.MSE) or (self.loss_type == LossType.RESCALED_MSE)): model_output = model(x_t, self._scale_timesteps(t), **model_kwargs) if (self.model_var_type in [ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE]): (B, C) = x_t.shape[:2] assert (model_output.shape == (B, (C * 2), *x_t.shape[2:])) (model_output, model_var_values) = th.split(model_output, C, dim=1) frozen_out = th.cat([model_output.detach(), model_var_values], dim=1) terms['vb'] = self._vb_terms_bpd(model=(lambda *args, r=frozen_out: r), x_start=x_start, x_t=x_t, t=t, clip_denoised=False)['output'] if (self.loss_type == LossType.RESCALED_MSE): terms['vb'] *= (self.num_timesteps / 1000.0) target = {ModelMeanType.PREVIOUS_X: self.q_posterior_mean_variance(x_start=x_start, x_t=x_t, t=t)[0], ModelMeanType.START_X: x_start, ModelMeanType.EPSILON: noise}[self.model_mean_type] assert (model_output.shape == target.shape == x_start.shape) terms['mse'] = mean_flat(((target - model_output) ** 2)) if ('vb' in terms): terms['loss'] = (terms['mse'] + terms['vb']) else: terms['loss'] = terms['mse'] else: raise NotImplementedError(self.loss_type) return terms def _prior_bpd(self, x_start): "\n Get the prior KL term for the variational lower-bound, measured in\n bits-per-dim.\n\n This term can't be optimized, as it only depends on the encoder.\n\n :param x_start: the [N x C x ...] tensor of inputs.\n :return: a batch of [N] KL values (in bits), one per batch element.\n " batch_size = x_start.shape[0] t = th.tensor(([(self.num_timesteps - 1)] * batch_size), device=x_start.device) (qt_mean, _, qt_log_variance) = self.q_mean_variance(x_start, t) kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0) return (mean_flat(kl_prior) / np.log(2.0)) def calc_bpd_loop(self, model, x_start, clip_denoised=True, model_kwargs=None): '\n Compute the entire variational lower-bound, measured in bits-per-dim,\n as well as other related quantities.\n\n :param model: the model to evaluate loss on.\n :param x_start: the [N x C x ...] tensor of inputs.\n :param clip_denoised: if True, clip denoised samples.\n :param model_kwargs: if not None, a dict of extra keyword arguments to\n pass to the model. This can be used for conditioning.\n\n :return: a dict containing the following keys:\n - total_bpd: the total variational lower-bound, per batch element.\n - prior_bpd: the prior term in the lower-bound.\n - vb: an [N x T] tensor of terms in the lower-bound.\n - xstart_mse: an [N x T] tensor of x_0 MSEs for each timestep.\n - mse: an [N x T] tensor of epsilon MSEs for each timestep.\n ' device = x_start.device batch_size = x_start.shape[0] vb = [] xstart_mse = [] mse = [] for t in list(range(self.num_timesteps))[::(- 1)]: t_batch = th.tensor(([t] * batch_size), device=device) noise = th.randn_like(x_start) x_t = self.q_sample(x_start=x_start, t=t_batch, noise=noise) with th.no_grad(): out = self._vb_terms_bpd(model, x_start=x_start, x_t=x_t, t=t_batch, clip_denoised=clip_denoised, model_kwargs=model_kwargs) vb.append(out['output']) xstart_mse.append(mean_flat(((out['pred_xstart'] - x_start) ** 2))) eps = self._predict_eps_from_xstart(x_t, t_batch, out['pred_xstart']) mse.append(mean_flat(((eps - noise) ** 2))) vb = th.stack(vb, dim=1) xstart_mse = th.stack(xstart_mse, dim=1) mse = th.stack(mse, dim=1) prior_bpd = self._prior_bpd(x_start) total_bpd = (vb.sum(dim=1) + prior_bpd) return {'total_bpd': total_bpd, 'prior_bpd': prior_bpd, 'vb': vb, 'xstart_mse': xstart_mse, 'mse': mse}
def _extract_into_tensor(arr, timesteps, broadcast_shape): '\n Extract values from a 1-D numpy array for a batch of indices.\n\n :param arr: the 1-D numpy array.\n :param timesteps: a tensor of indices into the array to extract.\n :param broadcast_shape: a larger shape of K dimensions with the batch\n dimension equal to the length of timesteps.\n :return: a tensor of shape [batch_size, 1, ...] where the shape has K dims.\n ' res = th.from_numpy(arr).to(device=timesteps.device)[timesteps].float() while (len(res.shape) < len(broadcast_shape)): res = res[(..., None)] return res.expand(broadcast_shape)