code
stringlengths
101
5.91M
_cache(None) def cuda_toolkit_available(): try: call(['nvcc'], stdout=DEVNULL, stderr=DEVNULL) return True except FileNotFoundError: return False
def random_normal(size=(1,), trunc_val=2.5, rnd_state=None): if (rnd_state is None): rnd_state = np.random len = np.array(size).prod() result = np.empty((len,), dtype=np.float32) for i in range(len): while True: x = rnd_state.normal() if ((x >= (- trunc_val)) and (x <= trunc_val)): break result[i] = (x / trunc_val) return result.reshape(size)
def create_custom_splits_for_experiments(task_path): data_keys = [i[:(- 4)] for i in subfiles(os.path.join(task_path, 'nnUNetData_plans_v2.1_2D_stage0'), join=False, suffix='npz')] existing_splits = os.path.join(task_path, 'splits_final.pkl') splits = load_pickle(existing_splits) splits = splits[:5] unique_a_only = np.unique([i.split('_')[0] for i in data_keys if (i.find('_A_') != (- 1))]) unique_b_only = np.unique([i.split('_')[0] for i in data_keys if (i.find('_B_') != (- 1))]) num_train_a = int(np.round((0.8 * len(unique_a_only)))) num_train_b = int(np.round((0.8 * len(unique_b_only)))) p = RandomState(1234) idx_a_train = p.choice(len(unique_a_only), num_train_a, replace=False) idx_b_train = p.choice(len(unique_b_only), num_train_b, replace=False) identifiers_a_train = [unique_a_only[i] for i in idx_a_train] identifiers_b_train = [unique_b_only[i] for i in idx_b_train] identifiers_a_val = [i for i in unique_a_only if (i not in identifiers_a_train)] identifiers_b_val = [i for i in unique_b_only if (i not in identifiers_b_train)] splits.append({'train': [i for i in data_keys if (i.split('_')[0] in identifiers_a_train)], 'val': ([i for i in data_keys if (i.split('_')[0] in identifiers_a_val)] + [i for i in data_keys if (i.split('_')[0] in identifiers_b_val)])}) splits.append({'train': [i for i in data_keys if (i.split('_')[0] in identifiers_b_train)], 'val': ([i for i in data_keys if (i.split('_')[0] in identifiers_a_val)] + [i for i in data_keys if (i.split('_')[0] in identifiers_b_val)])}) splits.append({'train': ([i for i in data_keys if (i.split('_')[0] in identifiers_b_train)] + [i for i in data_keys if (i.split('_')[0] in identifiers_a_train)]), 'val': ([i for i in data_keys if (i.split('_')[0] in identifiers_a_val)] + [i for i in data_keys if (i.split('_')[0] in identifiers_b_val)])}) save_pickle(splits, existing_splits)
def _config_dict_write(fp: IO[str], config: ConfigDict) -> None: fp.write(config.to_json(indent=4))
def get_rap_jitter(frequencies, p_floor, p_ceil, max_p_factor): counter = 0 cumsum = 0 mean_period = get_mean_period(frequencies, p_floor, p_ceil, max_p_factor) for (freq1, freq2, freq3) in shifted_sequence(frequencies, 3): if validate_frequencies([freq1, freq2, freq3], p_floor, p_ceil, max_p_factor): cumsum += np.abs(((1 / freq2) - ((((1 / freq1) + (1 / freq2)) + (1 / freq3)) / 3))) counter += 1 if (counter != 0): rap_jitter = (((cumsum / counter) / mean_period) if (mean_period != 0) else None) return rap_jitter return None
def main(args): models = [x[0] for x in args.model] tokenizer = AutoTokenizer.from_pretrained(models[0], model_max_length=sys.maxsize, padding_side='right', trust_remote_code=True) if args.fixed_length: lengths = [args.fixed_length] tokens = [len(tokenizer.encode(generate_prompt(args.fixed_length)[0]))] print(f'Prompt is {tokens[0]} tokens') else: if args.tokens_step: tokens = [x for x in range(args.min_tokens, (args.max_tokens + 1), args.tokens_step)] else: tokens = [args.min_tokens] while (args.min_tokens < args.max_tokens): point = (tokens[(- 1)] * 2) if (point <= args.max_tokens): tokens.append(point) else: break lengths = [] last_n = 0 for target in tqdm(tokens, desc='Determining sequence lengths'): num_tokens = 0 n = last_n while (num_tokens < target): last_n = n n += args.length_step prompt = generate_prompt(n)[0] num_tokens = len(tokenizer.encode(prompt)) lengths.append(last_n) results = [] for model in tqdm(models, desc='Model', leave=False): torch.cuda.empty_cache() loaded = load_model_and_apply_patches(model, args) pipe = pipeline('text-generation', model=loaded, tokenizer=tokenizer, pad_token_id=tokenizer.eos_token_id) result = ([0] * len(lengths)) for (i, length) in tenumerate(lengths, desc='Lengths', leave=False): for _ in trange(0, args.iterations, desc='Iterations', leave=False): (prompt_text, pass_key) = generate_prompt(length) num_tokens = len(pipe.tokenizer.encode(prompt_text)) answer = test_model(pipe, prompt_text, pass_key) if (answer == pass_key): result[i] += 1 result[i] /= args.iterations print(f'{model}: {tokens[i]}={int((result[i] * 100))}%') result.insert(0, model) results.append(result) if args.output_file: with open(args.output_file, 'w', encoding='utf-8') as f: f.write(f''',{','.join([str(x) for x in tokens])} ''') for result in results: f.write(f'''{','.join([str(x) for x in result])} ''')
def wind(cost, data, params, width): predicted_cp = [] for dataset in data: stsc = StandardScaler() signal = stsc.fit_transform(dataset.drop('changepoint', axis=1)) algo = rpt.Window(model=cost, params=params, width=width, jump=1) algo.fit(signal) my_bkps = algo.predict(n_bkps=len(dataset[(dataset['changepoint'] == 1)])) single_predicted_cp = pd.Series(data=0, index=dataset.index) single_predicted_cp[single_predicted_cp.index[my_bkps[:(- 1)]]] = 1 predicted_cp.append(single_predicted_cp) true_cp = [dataset.changepoint for dataset in data] nab = evaluating_change_point(true_cp, predicted_cp, metric='nab', numenta_time='30 sec') return nab
def main(): clorsv = input('client or server ? (type c or s) ') if (clorsv == 'c'): client() else: from phcpy import solver if (clorsv == 's'): nbsys = eval(input('-> how many systems ? ')) probs = [solver.random_trinomials() for i in range(0, nbsys)] nbclt = eval(input('-> how many clients ? ')) start_server(probs, nbclt) else: print('sorry, expected c or s, please try again')
def saliency_map_I_gradient(numpy_image, model, attr_objective, baseline='gaus', fold=10, interp='linear'): numpy_baseline = np.moveaxis((IG_baseline((np.moveaxis(numpy_image, 0, 2) * 255.0), mode=baseline) / 255.0), 2, 0) (grad_list, result_list, _) = I_gradient(numpy_image, numpy_baseline, model, attr_objective, fold, interp='linear') final_grad = (grad_list.mean(axis=0) * (numpy_image - numpy_baseline)) return (final_grad, result_list[(- 1)])
class UserAgent(object): def __init__(self, id: str, group=None, batch_size=0, learning_rate=0, beta=0, lamda=0, local_epochs=0, optimizer='sgd', K=None, personal_learning_rate=None, loss='xent', label_mode='supervised'): self.id = id (self.group, self.group_name) = group self.label_mode = label_mode self.batch_size = batch_size self.beta = beta self.lamda = lamda self.local_epochs = local_epochs if isinstance(optimizer, DictConfig): self.optimizer_name = optimizer.name self.learning_rate = optimizer.learning_rate self.personal_learning_rate = optimizer.personal_learning_rate self.optimizer_config = optimizer else: self.optimizer_name = optimizer self.learning_rate = learning_rate self.personal_learning_rate = personal_learning_rate self.K = K self.model = None
def load_state_dict(checkpoint_file: Union[(str, os.PathLike)], variant: Optional[str]=None): try: if (os.path.basename(checkpoint_file) == _add_variant(WEIGHTS_NAME, variant)): return torch.load(checkpoint_file, map_location='cpu') else: return safetensors.torch.load_file(checkpoint_file, device='cpu') except Exception as e: try: with open(checkpoint_file) as f: if f.read().startswith('version'): raise OSError('You seem to have cloned a repository without having git-lfs installed. Please install git-lfs and run `git lfs install` followed by `git lfs pull` in the folder you cloned.') else: raise ValueError(f'Unable to locate the file {checkpoint_file} which is necessary to load this pretrained model. Make sure you have saved the model properly.') from e except (UnicodeDecodeError, ValueError): raise OSError(f"Unable to load weights from checkpoint file for '{checkpoint_file}' at '{checkpoint_file}'. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.")
class STRes16UNetIN50(STRes16UNet50): NORM_TYPE = NormType.SPARSE_INSTANCE_NORM BLOCK = BottleneckIN
class MinitaurAlternatingLegsEnv(minitaur_gym_env.MinitaurGymEnv): metadata = {'render.modes': ['human', 'rgb_array'], 'video.frames_per_second': 66} def __init__(self, urdf_version=None, control_time_step=0.006, action_repeat=6, control_latency=0, pd_latency=0, on_rack=False, motor_kp=1.0, motor_kd=0.02, remove_default_joint_damping=False, render=False, num_steps_to_log=1000, env_randomizer=None, log_path=None): self._swing_offset = np.zeros(NUM_LEGS) self._extension_offset = np.zeros(NUM_LEGS) super(MinitaurAlternatingLegsEnv, self).__init__(urdf_version=urdf_version, accurate_motor_model_enabled=True, motor_overheat_protection=True, hard_reset=False, motor_kp=motor_kp, motor_kd=motor_kd, remove_default_joint_damping=remove_default_joint_damping, control_latency=control_latency, pd_latency=pd_latency, on_rack=on_rack, render=render, num_steps_to_log=num_steps_to_log, env_randomizer=env_randomizer, log_path=log_path, control_time_step=control_time_step, action_repeat=action_repeat) action_dim = 8 action_high = np.array(([0.1] * action_dim)) self.action_space = spaces.Box((- action_high), action_high) self._cam_dist = 1.0 self._cam_yaw = 30 self._cam_pitch = (- 30) def _reset(self): self.desired_pitch = DESIRED_PITCH init_pose = [(INIT_SWING_POS + self._swing_offset[0]), (INIT_SWING_POS + self._swing_offset[1]), (INIT_SWING_POS + self._swing_offset[2]), (INIT_SWING_POS + self._swing_offset[3]), (INIT_EXTENSION_POS + self._extension_offset[0]), (INIT_EXTENSION_POS + self._extension_offset[1]), (INIT_EXTENSION_POS + self._extension_offset[2]), (INIT_EXTENSION_POS + self._extension_offset[3])] initial_motor_angles = self._convert_from_leg_model(init_pose) super(MinitaurAlternatingLegsEnv, self)._reset(initial_motor_angles=initial_motor_angles, reset_duration=0.5) return self._get_observation() def _convert_from_leg_model(self, leg_pose): motor_pose = np.zeros(NUM_MOTORS) for i in range(NUM_LEGS): motor_pose[(2 * i)] = (leg_pose[(NUM_LEGS + i)] - (((- 1) ** (i / 2)) * leg_pose[i])) motor_pose[((2 * i) + 1)] = (leg_pose[(NUM_LEGS + i)] + (((- 1) ** (i / 2)) * leg_pose[i])) return motor_pose def _signal(self, t): initial_pose = np.array([INIT_SWING_POS, INIT_SWING_POS, INIT_SWING_POS, INIT_SWING_POS, INIT_EXTENSION_POS, INIT_EXTENSION_POS, INIT_EXTENSION_POS, INIT_EXTENSION_POS]) amplitude = STEP_AMPLITUDE period = STEP_PERIOD extension = (amplitude * ((- 1.0) + math.cos((((2 * math.pi) / period) * t)))) ith_leg = (int((t / period)) % 2) first_leg = np.array([0, 0, 0, 0, 0, extension, extension, 0]) second_leg = np.array([0, 0, 0, 0, extension, 0, 0, extension]) if ith_leg: signal = (initial_pose + second_leg) else: signal = (initial_pose + first_leg) return signal def _transform_action_to_motor_command(self, action): action[0:4] += self._swing_offset action[4:8] += self._extension_offset action += self._signal(self.minitaur.GetTimeSinceReset()) action = self._convert_from_leg_model(action) return action def is_fallen(self): orientation = self.minitaur.GetBaseOrientation() rot_mat = self._pybullet_client.getMatrixFromQuaternion(orientation) local_up = rot_mat[6:] return (np.dot(np.asarray([0, 0, 1]), np.asarray(local_up)) < 0.85) def _reward(self): return 1.0 def _get_true_observation(self): observation = [] (roll, pitch, _) = self.minitaur.GetTrueBaseRollPitchYaw() (roll_rate, pitch_rate, _) = self.minitaur.GetTrueBaseRollPitchYawRate() observation.extend([roll, pitch, roll_rate, pitch_rate]) observation[1] -= self.desired_pitch self._true_observation = np.array(observation) return self._true_observation def _get_observation(self): observation = [] (roll, pitch, _) = self.minitaur.GetBaseRollPitchYaw() (roll_rate, pitch_rate, _) = self.minitaur.GetBaseRollPitchYawRate() observation.extend([roll, pitch, roll_rate, pitch_rate]) observation[1] -= self.desired_pitch self._observation = np.array(observation) return self._observation def _get_observation_upper_bound(self): upper_bound = np.zeros(self._get_observation_dimension()) upper_bound[0:2] = (2 * math.pi) upper_bound[2:4] = ((2 * math.pi) / self._time_step) return upper_bound def _get_observation_lower_bound(self): lower_bound = (- self._get_observation_upper_bound()) return lower_bound def set_swing_offset(self, value): self._swing_offset = value def set_extension_offset(self, value): self._extension_offset = value def set_desired_pitch(self, value): self.desired_pitch = value
def drop_table(): db = database() sql_script = '\n DROP TABLE `class`;\n DROP TABLE `method`;\n ' db.execute(sql_script)
def load_tokens(filename): with open(filename) as infile: data = json.load(infile) tokens = [] for d in data: tokens += d['token'] print('{} tokens from {} examples loaded from {}.'.format(len(tokens), len(data), filename)) return tokens
class PolyBUnit(PolyBaseUnit): def __init__(self, two_way_scale, poly_scale): super(PolyBUnit, self).__init__(two_way_scale=two_way_scale, two_way_block=TwoWayBBlock, poly_scale=poly_scale, poly_res_block=poly_res_b_block, poly_pre_block=PolyPreBBlock)
class AuxiliaryLossAverageMeter(object): def __init__(self): self.aux_loss = {} def update(self, value_dict, accu_steps, n=1): if (value_dict is not None): for (key, value) in value_dict.items(): if (key not in self.aux_loss): self.aux_loss[key] = AverageMeter() self.aux_loss[key].update((value / accu_steps), n) else: self.aux_loss[key].update((value / accu_steps), n) def reset(self): if (len(self.aux_loss) > 0): for key in self.aux_loss.keys(): self.aux_loss[key].reset() def avg_infos(self): if (len(self.aux_loss) == 0): return '' else: infos = [] for key in self.aux_loss.keys(): infos.append(('%s: %.3f' % (key, self.aux_loss[key].avg))) return (', ' + ', '.join(infos))
def load_state_from_resource(model, resource: str): print(f'Using pretrained resource {resource}') if resource.startswith('url::'): url = resource.split('url::')[1] return load_state_dict_from_url(model, url, progress=True) elif resource.startswith('local::'): path = resource.split('local::')[1] return load_wts(model, path) else: raise ValueError('Invalid resource type, only url:: and local:: are supported')
class BartTokenizer(PreTrainedTokenizer): vocab_files_names = VOCAB_FILES_NAMES pretrained_vocab_files_map = PRETRAINED_VOCAB_FILES_MAP max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES model_input_names = ['input_ids', 'attention_mask'] def __init__(self, vocab_file, merges_file, errors='replace', bos_token='<s>', eos_token='</s>', sep_token='</s>', cls_token='<s>', unk_token='<unk>', pad_token='<pad>', mask_token='<mask>', add_prefix_space=False, **kwargs): bos_token = (AddedToken(bos_token, lstrip=False, rstrip=False) if isinstance(bos_token, str) else bos_token) eos_token = (AddedToken(eos_token, lstrip=False, rstrip=False) if isinstance(eos_token, str) else eos_token) sep_token = (AddedToken(sep_token, lstrip=False, rstrip=False) if isinstance(sep_token, str) else sep_token) cls_token = (AddedToken(cls_token, lstrip=False, rstrip=False) if isinstance(cls_token, str) else cls_token) unk_token = (AddedToken(unk_token, lstrip=False, rstrip=False) if isinstance(unk_token, str) else unk_token) pad_token = (AddedToken(pad_token, lstrip=False, rstrip=False) if isinstance(pad_token, str) else pad_token) mask_token = (AddedToken(mask_token, lstrip=True, rstrip=False) if isinstance(mask_token, str) else mask_token) super().__init__(errors=errors, bos_token=bos_token, eos_token=eos_token, unk_token=unk_token, sep_token=sep_token, cls_token=cls_token, pad_token=pad_token, mask_token=mask_token, add_prefix_space=add_prefix_space, **kwargs) with open(vocab_file, encoding='utf-8') as vocab_handle: self.encoder = json.load(vocab_handle) self.decoder = {v: k for (k, v) in self.encoder.items()} self.errors = errors self.byte_encoder = bytes_to_unicode() self.byte_decoder = {v: k for (k, v) in self.byte_encoder.items()} with open(merges_file, encoding='utf-8') as merges_handle: bpe_merges = merges_handle.read().split('\n')[1:(- 1)] bpe_merges = [tuple(merge.split()) for merge in bpe_merges] self.bpe_ranks = dict(zip(bpe_merges, range(len(bpe_merges)))) self.cache = {} self.add_prefix_space = add_prefix_space self.pat = re.compile("'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)|\\s+") def vocab_size(self): return len(self.encoder) def get_vocab(self): return dict(self.encoder, **self.added_tokens_encoder) def bpe(self, token): if (token in self.cache): return self.cache[token] word = tuple(token) pairs = get_pairs(word) if (not pairs): return token 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) except ValueError: new_word.extend(word[i:]) break else: new_word.extend(word[i:j]) i = j 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 _tokenize(self, text): bpe_tokens = [] for token in re.findall(self.pat, text): token = ''.join((self.byte_encoder[b] for b in token.encode('utf-8'))) bpe_tokens.extend((bpe_token for bpe_token in self.bpe(token).split(' '))) return bpe_tokens def _convert_token_to_id(self, token): return self.encoder.get(token, self.encoder.get(self.unk_token)) def _convert_id_to_token(self, index): return self.decoder.get(index) def convert_tokens_to_string(self, tokens): text = ''.join(tokens) text = bytearray([self.byte_decoder[c] for c in text]).decode('utf-8', errors=self.errors) return text def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str]=None) -> Tuple[str]: if (not os.path.isdir(save_directory)): logger.error(f'Vocabulary path ({save_directory}) should be a directory') return vocab_file = os.path.join(save_directory, (((filename_prefix + '-') if filename_prefix else '') + VOCAB_FILES_NAMES['vocab_file'])) merge_file = os.path.join(save_directory, (((filename_prefix + '-') if filename_prefix else '') + VOCAB_FILES_NAMES['merges_file'])) with open(vocab_file, 'w', encoding='utf-8') as f: f.write((json.dumps(self.encoder, indent=2, sort_keys=True, ensure_ascii=False) + '\n')) index = 0 with open(merge_file, 'w', encoding='utf-8') as writer: writer.write('#version: 0.2\n') for (bpe_tokens, token_index) in sorted(self.bpe_ranks.items(), key=(lambda kv: kv[1])): if (index != token_index): logger.warning(f'Saving vocabulary to {merge_file}: BPE merge indices are not consecutive. Please check that the tokenizer is not corrupted!') index = token_index writer.write((' '.join(bpe_tokens) + '\n')) index += 1 return (vocab_file, merge_file) def build_inputs_with_special_tokens(self, token_ids_0: List[int], token_ids_1: Optional[List[int]]=None) -> List[int]: if (token_ids_1 is None): return (([self.cls_token_id] + token_ids_0) + [self.sep_token_id]) cls = [self.cls_token_id] sep = [self.sep_token_id] return (((((cls + token_ids_0) + sep) + sep) + token_ids_1) + sep) def get_special_tokens_mask(self, token_ids_0: List[int], token_ids_1: Optional[List[int]]=None, already_has_special_tokens: bool=False) -> List[int]: if already_has_special_tokens: return super().get_special_tokens_mask(token_ids_0=token_ids_0, token_ids_1=token_ids_1, already_has_special_tokens=True) if (token_ids_1 is None): return (([1] + ([0] * len(token_ids_0))) + [1]) return (((([1] + ([0] * len(token_ids_0))) + [1, 1]) + ([0] * len(token_ids_1))) + [1]) def create_token_type_ids_from_sequences(self, token_ids_0: List[int], token_ids_1: Optional[List[int]]=None) -> List[int]: sep = [self.sep_token_id] cls = [self.cls_token_id] if (token_ids_1 is None): return (len(((cls + token_ids_0) + sep)) * [0]) return (len((((((cls + token_ids_0) + sep) + sep) + token_ids_1) + sep)) * [0]) def prepare_for_tokenization(self, text, is_split_into_words=False, **kwargs): add_prefix_space = kwargs.pop('add_prefix_space', self.add_prefix_space) if ((is_split_into_words or add_prefix_space) and ((len(text) > 0) and (not text[0].isspace()))): text = (' ' + text) return (text, kwargs)
def get_new_pallete(num_cls): n = num_cls pallete = ([0] * (n * 3)) for j in range(0, n): lab = j pallete[((j * 3) + 0)] = 0 pallete[((j * 3) + 1)] = 0 pallete[((j * 3) + 2)] = 0 i = 0 while (lab > 0): pallete[((j * 3) + 0)] |= (((lab >> 0) & 1) << (7 - i)) pallete[((j * 3) + 1)] |= (((lab >> 1) & 1) << (7 - i)) pallete[((j * 3) + 2)] |= (((lab >> 2) & 1) << (7 - i)) i = (i + 1) lab >>= 3 return pallete
class TimeSeriesData(): def __init__(self, variable: Iterable=None, start_time_step: int=None, end_time_step: int=None): self.variable = (variable if (variable is None) else np.array(variable)) self.start_time_step = start_time_step self.end_time_step = end_time_step def __getattr__(self, name: str, start_time_step: int=None, end_time_step: int=None): try: variable = self.__dict__[f'_{name}'] except KeyError: raise AttributeError(f'_{name}') if isinstance(variable, Iterable): start_time_step = (self.start_time_step if (start_time_step is None) else start_time_step) start_index = (0 if (start_time_step is None) else start_time_step) end_time_step = (self.end_time_step if (end_time_step is None) else end_time_step) end_index = (len(variable) if (end_time_step is None) else (end_time_step + 1)) return variable[start_index:end_index] else: return variable def __setattr__(self, name: str, value: Any): self.__dict__[f'_{name}'] = value
class MMapIndexedDatasetBuilder(object): def __init__(self, out_file, dtype=np.int64): self._data_file = open(out_file, 'wb') self._dtype = dtype self._sizes = [] def add_item(self, tensor): np_array = np.array(tensor.numpy(), dtype=self._dtype) self._data_file.write(np_array.tobytes(order='C')) self._sizes.append(np_array.size) def merge_file_(self, another_file): index = MMapIndexedDataset.Index(index_file_path(another_file)) assert (index.dtype == self._dtype) for size in index.sizes: self._sizes.append(size) with open(data_file_path(another_file), 'rb') as f: shutil.copyfileobj(f, self._data_file) def finalize(self, index_file): self._data_file.close() with MMapIndexedDataset.Index.writer(index_file, self._dtype) as index: index.write(self._sizes)
def _download_bsd300(dest='dataset'): output_image_dir = join(dest, 'BSDS300/images') if (not exists(output_image_dir)): makedirs(dest, exist_ok=True) url = ' print('downloading url ', url) data = urllib.request.urlopen(url) file_path = join(dest, basename(url)) with open(file_path, 'wb') as f: f.write(data.read()) print('Extracting data') with tarfile.open(file_path) as tar: for item in tar: tar.extract(item, dest) remove(file_path) return output_image_dir
def split_dataset(filenames, train_set, val_set, test_set, val_len=None, test_len=None): len_filenames = len(filenames) len_all_sets = ((len(train_set) + len(val_set)) + len(test_set)) files_added = (len_filenames - len_all_sets) print('Total {} files, with {} files already in txt files.'.format(len_filenames, len_all_sets)) print('{} files added'.format(files_added)) if (files_added is 0): print('No need to split. Returning original results.') return (train_set, val_set, test_set) if (files_added < 0): raise Exception('split_dataset: Total file count is smaller than combined length of train/val/test set!') if ((len(test_set) is 0) and (test_len is None)): raise ValueError('split_dataset: test_set is empty and no test_len is specified!') if ((len(val_set) is 0) and (val_len is None)): raise ValueError('split_dataset: val_set is empty and no val_len is specified!') if (test_len is None): test_len = len(test_set) if (val_len is None): val_len = len(val_set) if ((len(val_set) is 0) and (len(test_set) is 0)): if (len(train_set) != 0): not_train_set = [filename for filename in filenames if (filename not in train_set)] else: not_train_set = filenames val_set = not_train_set[0:val_len] test_set = not_train_set[val_len:(val_len + test_len)] train_set += not_train_set[(val_len + test_len):] else: not_val_or_test_set = [filename for filename in filenames if ((filename not in val_set) and (filename not in test_set))] len_diff = (val_len - len(val_set)) if (len_diff > 0): val_set += not_val_or_test_set[0:len_diff] not_val_or_test_set = not_val_or_test_set[len_diff:] print('Expected val set length: {}, current val set length: {}'.format(val_len, len(val_set))) print(('Added %d files to val set.' % len_diff)) print('Unusual behavior. Do you really want to add files to val set?') elif (len_diff < 0): print('Expected val set length: {}, current val set length: {}'.format(val_len, len(val_set))) raise RuntimeError('split_dataset: Expected val length is smaller than current length!') len_diff = (test_len - len(test_set)) if (len_diff > 0): test_set += not_val_or_test_set[0:len_diff] not_val_or_test_set = not_val_or_test_set[len_diff:] print('Expected test set length: {}, current test set length: {}'.format(val_len, len(val_set))) print(('Added %d files to test set.' % len_diff)) print('Unusual behavior. Do you really want to add files to test set?') elif (len_diff < 0): print('Expected test set length: {}, current test set length: {}'.format(val_len, len(val_set))) raise RuntimeError('split_dataset: Expected test length is smaller than current length!') train_set = not_val_or_test_set return (train_set, val_set, test_set)
class ScaledDotProductAttention(nn.Module): def __init__(self, temperature, attn_dropout=0.1): super().__init__() self.temperature = temperature self.dropout = nn.Dropout(attn_dropout) self.softmax = nn.Softmax(dim=2) def forward(self, q, k, v): attn = torch.bmm(q, k.transpose(1, 2)) attn = (attn / self.temperature) log_attn = F.log_softmax(attn, 2) attn = self.softmax(attn) attn = self.dropout(attn) output = torch.bmm(attn, v) return (output, attn, log_attn)
def build_pose_evaluator(args): classes_path = (args.dataset_path + args.class_info) classes = load_classes(classes_path) models_path = (args.dataset_path + args.models) (models, models_info) = load_models(models_path, classes) symmetries_path = (args.dataset_path + args.model_symmetry) model_symmetry = load_model_symmetry(symmetries_path, classes) classes = [classes[k] for k in classes] if (args.dataset == 'ycbv'): evaluator = PoseEvaluator(models, classes, models_info, model_symmetry) elif (args.dataset == 'lmo'): evaluator = PoseEvaluatorLMO(models, classes, models_info, model_symmetry) else: raise ValueError('Unknown dataset.') return evaluator
def load_flax_weights_in_pytorch_model(pt_model, flax_state): try: import torch except ImportError: logger.error('Loading a Flax weights in PyTorch, requires both PyTorch and Flax to be installed. Please see and for installation instructions.') raise is_type_bf16 = flatten_dict(jax.tree_map((lambda x: (x.dtype == jnp.bfloat16)), flax_state)).values() if any(is_type_bf16): logger.warning('Found ``bfloat16`` weights in Flax model. Casting all ``bfloat16`` weights to ``float32`` before loading those in PyTorch model.') flax_state = jax.tree_map((lambda params: (params.astype(np.float32) if (params.dtype == jnp.bfloat16) else params)), flax_state) flax_state_dict = flatten_dict(flax_state) pt_model_dict = pt_model.state_dict() load_model_with_head_into_base_model = ((pt_model.base_model_prefix in flax_state) and (pt_model.base_model_prefix not in set([k.split('.')[0] for k in pt_model_dict.keys()]))) load_base_model_into_model_with_head = ((pt_model.base_model_prefix not in flax_state) and (pt_model.base_model_prefix in set([k.split('.')[0] for k in pt_model_dict.keys()]))) unexpected_keys = [] missing_keys = set(pt_model_dict.keys()) for (flax_key_tuple, flax_tensor) in flax_state_dict.items(): has_base_model_prefix = (flax_key_tuple[0] == pt_model.base_model_prefix) require_base_model_prefix = ('.'.join(((pt_model.base_model_prefix,) + flax_key_tuple)) in pt_model_dict) if (load_model_with_head_into_base_model and has_base_model_prefix): flax_key_tuple = flax_key_tuple[1:] elif (load_base_model_into_model_with_head and require_base_model_prefix): flax_key_tuple = ((pt_model.base_model_prefix,) + flax_key_tuple) if ((flax_key_tuple[(- 1)] == 'kernel') and (flax_tensor.ndim == 4) and ('.'.join(flax_key_tuple) not in pt_model_dict)): flax_key_tuple = (flax_key_tuple[:(- 1)] + ('weight',)) flax_tensor = jnp.transpose(flax_tensor, (3, 2, 0, 1)) elif ((flax_key_tuple[(- 1)] == 'kernel') and ('.'.join(flax_key_tuple) not in pt_model_dict)): flax_key_tuple = (flax_key_tuple[:(- 1)] + ('weight',)) flax_tensor = flax_tensor.T elif (flax_key_tuple[(- 1)] in ['scale', 'embedding']): flax_key_tuple = (flax_key_tuple[:(- 1)] + ('weight',)) flax_key = '.'.join(flax_key_tuple) if (flax_key in pt_model_dict): if (flax_tensor.shape != pt_model_dict[flax_key].shape): raise ValueError(f'Flax checkpoint seems to be incorrect. Weight {flax_key_tuple} was expected to be of shape {pt_model_dict[flax_key].shape}, but is {flax_tensor.shape}.') else: flax_tensor = (np.asarray(flax_tensor) if (not isinstance(flax_tensor, np.ndarray)) else flax_tensor) pt_model_dict[flax_key] = torch.from_numpy(flax_tensor) missing_keys.remove(flax_key) else: unexpected_keys.append(flax_key) pt_model.load_state_dict(pt_model_dict) missing_keys = list(missing_keys) if (len(unexpected_keys) > 0): logger.warning(f'''Some weights of the Flax model were not used when initializing the PyTorch model {pt_model.__class__.__name__}: {unexpected_keys} - This IS expected if you are initializing {pt_model.__class__.__name__} from a Flax model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a FlaxBertForPreTraining model). - This IS NOT expected if you are initializing {pt_model.__class__.__name__} from a Flax model that you expect to be exactly identical (e.g. initializing a BertForSequenceClassification model from a FlaxBertForSequenceClassification model).''') else: logger.warning(f'''All Flax model weights were used when initializing {pt_model.__class__.__name__}. ''') if (len(missing_keys) > 0): logger.warning(f'''Some weights of {pt_model.__class__.__name__} were not initialized from the Flax model and are newly initialized: {missing_keys} You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.''') else: logger.warning(f'''All the weights of {pt_model.__class__.__name__} were initialized from the Flax model. If your task is similar to the task the model of the checkpoint was trained on, you can already use {pt_model.__class__.__name__} for predictions without further training.''') return pt_model
class LookUpTable(object): def __init__(self, objlist): self.idx2obj = dict(enumerate(objlist)) self.obj2idx = {v: k for (k, v) in six.iteritems(self.idx2obj)} def size(self): return len(self.idx2obj) def get_obj(self, idx): return self.idx2obj[idx] def get_idx(self, obj): return self.obj2idx[obj] def __str__(self): return self.idx2obj.__str__()
class Morphology(lazylist): def __init__(self, path='', known={}): self.known = known self._path = path self._cmd = set(('word', 'char', 'haspref', 'hassuf', 'addpref', 'addsuf', 'deletepref', 'deletesuf', 'goodleft', 'goodright')) self._cmd.update([('f' + x) for x in self._cmd]) def path(self): return self._path def load(self): list.extend(self, (x.split() for x in _read(self._path))) def apply(self, token, previous=(None, None), next=(None, None)): w = token[0] for r in self: if (r[1] in self._cmd): (f, x, pos, cmd) = (bool(0), r[0], r[(- 2)], r[1].lower()) if (r[2] in self._cmd): (f, x, pos, cmd) = (bool(1), r[1], r[(- 2)], r[2].lower().lstrip('f')) if (f and (token[1] != r[0])): continue if (((cmd == 'word') and (x == w)) or ((cmd == 'char') and (x in w)) or ((cmd == 'haspref') and w.startswith(x)) or ((cmd == 'hassuf') and w.endswith(x)) or ((cmd == 'addpref') and ((x + w) in self.known)) or ((cmd == 'addsuf') and ((w + x) in self.known)) or ((cmd == 'deletepref') and w.startswith(x) and (w[len(x):] in self.known)) or ((cmd == 'deletesuf') and w.endswith(x) and (w[:(- len(x))] in self.known)) or ((cmd == 'goodleft') and (x == next[0])) or ((cmd == 'goodright') and (x == previous[0]))): token[1] = pos return token def insert(self, i, tag, affix, cmd='hassuf', tagged=None): if (affix.startswith('-') and affix.endswith('-')): (affix, cmd) = (affix[(+ 1):(- 1)], 'char') if affix.startswith('-'): (affix, cmd) = (affix[(+ 1):(- 0)], 'hassuf') if affix.endswith('-'): (affix, cmd) = (affix[(+ 0):(- 1)], 'haspref') if tagged: r = [tagged, affix, ('f' + cmd.lstrip('f')), tag, 'x'] else: r = [affix, cmd.lstrip('f'), tag, 'x'] lazylist.insert(self, i, r) def append(self, *args, **kwargs): self.insert((len(self) - 1), *args, **kwargs) def extend(self, rules=[]): for r in rules: self.append(*r)
def process(stride, path, scene_name): ds = Dataset() ds.load((path + scene_name)) features = [] labels = [] processor = IrlDataProcessor(ds) scene = ds.get('scene', ds.list_scenes()[0]) frame_list = [] frame_token = scene['first_frame'] while frame_token: frame_list.append(frame_token) frame = ds.get('frame', frame_token) frame_token = frame['next'] for frame_idx in tqdm(range(0, len(frame_list), stride)): frame_token = frame_list[frame_idx] frame = ds.get('frame', frame_token) for inst_token in frame['instances']: instance = ds.get('instance', inst_token) agent = ds.get('agent', instance['agent_token']) if (agent['type'] not in {'Pedestrian', 'Undefined'}): try: if (ds.get_inst_mode(inst_token) != 'incoming'): continue spot_centers = processor.detect_center(inst_token, 'spot') if (spot_centers == []): continue agent_centers = processor.detect_center(inst_token, 'agent') thres = 3.5 ego_speed = instance['speed'] feature = np.zeros((NUM_FEATURE, len(spot_centers))) for (center_idx, center_coords) in enumerate(spot_centers): local_offset = processor.compute_relative_offset(inst_token, center_coords) (astar_dist, astar_dir, astar_graph) = processor.compute_Astar_dist_dir(inst_token, center_coords) nearby_agents = 0 for center_agent in agent_centers: dist = astar_graph.dist_to_graph(center_agent) if (dist < thres): nearby_agents += 1 feature[(0, center_idx)] = local_offset[0] feature[(1, center_idx)] = local_offset[1] feature[(2, center_idx)] = astar_dist feature[(3, center_idx)] = astar_dir feature[(4, center_idx)] = nearby_agents feature[(5, center_idx)] = ego_speed label = processor.get_intent_label(inst_token, spot_centers) features.append(feature) labels.append(label) except Exception as err: print(('\nError occured for instance %s' % inst_token)) traceback.print_exc() frame_token = frame['next'] if (not os.path.exists(DATA_PATH)): os.mkdir(DATA_PATH) with open((DATA_PATH + ('/%s_feature.pkl' % scene_name)), 'wb') as f: pickle.dump(features, f) with open((DATA_PATH + ('/%s_label.pkl' % scene_name)), 'wb') as f: pickle.dump(labels, f)
class RelaxedOneHotCategoricalStraightThrough2D(RelaxedOneHotCategorical2D): event_dim = 3 def rsample(self, sample_shape=torch.Size()): soft_sample = super().rsample(sample_shape) soft_sample = clamp_probs(soft_sample) hard_sample = QuantizeCategorical2D.apply(soft_sample) return hard_sample def log_prob(self, value): value = getattr(value, '_unquantize', value) return super().log_prob(value)
class Boxban_Env1(BoxobanEnv): metadata = {'render.modes': ['human', 'rgb_array', 'tiny_human', 'tiny_rgb_array']} def __init__(self): super(Boxban_Env1, self).__init__(max_steps=200, difficulty='medium')
class DinoV2Features(TorchFeatureExtractor): tag = 'dinov2' license = 'Apache-2.0' citation = '\{oquab2023dinov2,\n title={DINOv2: Learning Robust Visual Features without Supervision},\n author={Oquab, Maxime and Darcet, Timothee and Moutakanni, Theo and Vo, Huy V. and Szafraniec, Marc and Khalidov, Vasil and Fernandez, Pierre and Haziza, Daniel and Massa, Francisco and El-Nouby, Alaaeldin and Howes, Russell and Huang, Po-Yao and Xu, Hu and Sharma, Vasu and Li, Shang-Wen and Galuba, Wojciech and Rabbat, Mike and Assran, Mido and Ballas, Nicolas and Synnaeve, Gabriel and Misra, Ishan and Jegou, Herve and Mairal, Julien and Labatut, Patrick and Joulin, Armand and Bojanowski, Piotr},\n journal={arXiv:2304.07193},\n year={2023}\n}\n' def __init__(self, cfg, weights, device=None, center_crop=False): super().__init__() self.cfg = cfg self.weights = weights self.device = torch_utils.get_device(device) self.model = build_model_for_eval(OmegaConf.load(cfg), weights) self.model.to(self.device) self.model.eval() self.num_features = 1024 all_transforms = ([transforms.CenterCrop(224)] if center_crop else []) all_transforms += [transforms.Lambda((lambda x: (x / 255.0))), transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))] self.transform = transforms.Compose(all_transforms) self.preprocess_kwargs = dict(standardize=False) self._center_crop = center_crop def dump_config(self): cls_name = self.__class__.__name__ return {'class': f'slideflow.model.extractors.dinov2.{cls_name}', 'kwargs': {'center_crop': self._center_crop, 'cfg': self.cfg, 'weights': self.weights}}
class BackEnd(Model): def __init__(self, half_res=False): super(BackEnd, self).__init__() self.half_res = half_res self.upsample = layers.UpSampling2D(2, interpolation='bilinear') self.conv1 = BaseConv(256, 1, 1, activation='relu', use_bn=True) self.conv2 = BaseConv(256, 3, 1, activation='relu', use_bn=True) self.conv3 = BaseConv(128, 1, 1, activation='relu', use_bn=True) self.conv4 = BaseConv(128, 3, 1, activation='relu', use_bn=True) self.conv5 = BaseConv(64, 1, 1, activation='relu', use_bn=True) self.conv6 = BaseConv(64, 3, 1, activation='relu', use_bn=True) self.conv7 = BaseConv(32, 3, 1, activation='relu', use_bn=True) if (not self.half_res): self.conv8 = BaseConv(32, 1, 1, activation='relu', use_bn=True) self.conv9 = BaseConv(32, 3, 1, activation='relu', use_bn=True) self.conv10 = BaseConv(32, 3, 1, activation='relu', use_bn=True) def call(self, inputs): if self.half_res: (conv2_2, conv3_3, conv4_3, conv5_3) = inputs else: (conv1_2, conv2_2, conv3_3, conv4_3, conv5_3) = inputs x = self.upsample(conv5_3) x = tf.concat([x, conv4_3], axis=(- 1)) x = self.conv1(x) x = self.conv2(x) x = self.upsample(x) x = tf.concat([x, conv3_3], axis=(- 1)) x = self.conv3(x) x = self.conv4(x) x = self.upsample(x) x = tf.concat([x, conv2_2], axis=(- 1)) x = self.conv5(x) x = self.conv6(x) x = self.conv7(x) if (not self.half_res): x = self.upsample(x) x = tf.concat([x, conv1_2], axis=(- 1)) x = self.conv8(x) x = self.conv9(x) x = self.conv10(x) return x
class DepthFirstSearch(AbstractSearch): def __init__(self, policies): self.policies = policies def __call__(self, root, max_expansions=10, *args, **kwargs): start_time = timeit.default_timer() self.policies.initialize(root) nodes_to_visit = [root] best_reward = (- float('inf')) best_node = None while (len(nodes_to_visit) > 0): node = nodes_to_visit.pop() if node.terminal: if (node.accumulatedReward() > best_reward): best_reward = node.accumulatedReward() best_node = node else: expanded = 0 for child in node.children: expanded += 1 self.policies.instantiate(node, child) nodes_to_visit.append(child) if (expanded > max_expansions): break path = [] if (best_node is not None): path.append(best_node) while (best_node.parent is not None): best_node = best_node.parent path.reverse() elapsed = (timeit.default_timer() - start_time) return (elapsed, path)
def mobilenet_v2(pretrained=False, progress=True, quantize=False, **kwargs): model = QuantizableMobileNetV2(block=QuantizableInvertedResidual, **kwargs) _replace_relu(model) if quantize: backend = 'qnnpack' quantize_model(model, backend) else: assert (pretrained in [True, False]) if pretrained: if quantize: model_url = quant_model_urls[('mobilenet_v2_' + backend)] else: model_url = model_urls['mobilenet_v2'] state_dict = load_state_dict_from_url(model_url, progress=progress) model.load_state_dict(state_dict) return model
class ClMetrics(_FullMetric): def __init__(self, **kwargs): super(ClMetrics, self).__init__(**kwargs) self.metrics = [average.MeanAccuracy(name='mean_accu_out')]
def qresnet18(pretrained=None, num_classes=1000): model = QResNet(block=BasicBlock, layers=[2, 2, 2, 2], num_classes=num_classes) if pretrained: ch = torch.load(pretrained) ch = {n.replace('module.', ''): v for (n, v) in ch['state_dict'].items()} model.load_state_dict(ch) print('loading pretrained') return model
class SigmoidFocalClassificationLossTest(tf.test.TestCase): def testEasyExamplesProduceSmallLossComparedToSigmoidXEntropy(self): prediction_tensor = tf.constant([[[_logit(0.97)], [_logit(0.9)], [_logit(0.73)], [_logit(0.27)], [_logit(0.09)], [_logit(0.03)]]], tf.float32) target_tensor = tf.constant([[[1], [1], [1], [0], [0], [0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, gamma=2.0, alpha=None) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) order_of_ratio = np.power(10, np.floor(np.log10((sigmoid_loss / focal_loss)))) self.assertAllClose(order_of_ratio, [[1000, 100, 10, 10, 100, 1000]]) def testHardExamplesProduceLossComparableToSigmoidXEntropy(self): prediction_tensor = tf.constant([[[_logit(0.55)], [_logit(0.52)], [_logit(0.5)], [_logit(0.48)], [_logit(0.45)]]], tf.float32) target_tensor = tf.constant([[[1], [1], [1], [0], [0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, gamma=2.0, alpha=None) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) order_of_ratio = np.power(10, np.floor(np.log10((sigmoid_loss / focal_loss)))) self.assertAllClose(order_of_ratio, [[1.0, 1.0, 1.0, 1.0, 1.0]]) def testNonAnchorWiseOutputComparableToSigmoidXEntropy(self): prediction_tensor = tf.constant([[[_logit(0.55)], [_logit(0.52)], [_logit(0.5)], [_logit(0.48)], [_logit(0.45)]]], tf.float32) target_tensor = tf.constant([[[1], [1], [1], [0], [0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=False, gamma=2.0, alpha=None) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=False) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) order_of_ratio = np.power(10, np.floor(np.log10((sigmoid_loss / focal_loss)))) self.assertAlmostEqual(order_of_ratio, 1.0) def testIgnoreNegativeExampleLossViaAlphaMultiplier(self): prediction_tensor = tf.constant([[[_logit(0.55)], [_logit(0.52)], [_logit(0.5)], [_logit(0.48)], [_logit(0.45)]]], tf.float32) target_tensor = tf.constant([[[1], [1], [1], [0], [0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, gamma=2.0, alpha=1.0) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) self.assertAllClose(focal_loss[0][3:], [0.0, 0.0]) order_of_ratio = np.power(10, np.floor(np.log10((sigmoid_loss[0][:3] / focal_loss[0][:3])))) self.assertAllClose(order_of_ratio, [1.0, 1.0, 1.0]) def testIgnorePositiveExampleLossViaAlphaMultiplier(self): prediction_tensor = tf.constant([[[_logit(0.55)], [_logit(0.52)], [_logit(0.5)], [_logit(0.48)], [_logit(0.45)]]], tf.float32) target_tensor = tf.constant([[[1], [1], [1], [0], [0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, gamma=2.0, alpha=0.0) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) self.assertAllClose(focal_loss[0][:3], [0.0, 0.0, 0.0]) order_of_ratio = np.power(10, np.floor(np.log10((sigmoid_loss[0][3:] / focal_loss[0][3:])))) self.assertAllClose(order_of_ratio, [1.0, 1.0]) def testSimilarToSigmoidXEntropyWithHalfAlphaAndZeroGammaUpToAScale(self): prediction_tensor = tf.constant([[[(- 100), 100, (- 100)], [100, (- 100), (- 100)], [100, 0, (- 100)], [(- 100), (- 100), 100]], [[(- 100), 0, 100], [(- 100), 100, (- 100)], [100, 100, 100], [0, 0, (- 1)]]], tf.float32) target_tensor = tf.constant([[[0, 1, 0], [1, 0, 0], [1, 0, 0], [0, 0, 1]], [[0, 0, 1], [0, 1, 0], [1, 1, 1], [1, 0, 0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1], [1, 1, 1, 0]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, alpha=0.5, gamma=0.0) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) self.assertAllClose(sigmoid_loss, (focal_loss * 2)) def testSameAsSigmoidXEntropyWithNoAlphaAndZeroGamma(self): prediction_tensor = tf.constant([[[(- 100), 100, (- 100)], [100, (- 100), (- 100)], [100, 0, (- 100)], [(- 100), (- 100), 100]], [[(- 100), 0, 100], [(- 100), 100, (- 100)], [100, 100, 100], [0, 0, (- 1)]]], tf.float32) target_tensor = tf.constant([[[0, 1, 0], [1, 0, 0], [1, 0, 0], [0, 0, 1]], [[0, 0, 1], [0, 1, 0], [1, 1, 1], [1, 0, 0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1], [1, 1, 1, 0]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=True, alpha=None, gamma=0.0) sigmoid_loss_op = losses.WeightedSigmoidClassificationLoss(anchorwise_output=True) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) sigmoid_loss = sigmoid_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: (sigmoid_loss, focal_loss) = sess.run([sigmoid_loss, focal_loss]) self.assertAllClose(sigmoid_loss, focal_loss) def testExpectedLossWithAlphaOneAndZeroGamma(self): prediction_tensor = tf.constant([[[0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0]], [[0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0]]], tf.float32) target_tensor = tf.constant([[[0, 1, 0], [1, 0, 0], [1, 0, 0], [0, 0, 1]], [[0, 0, 1], [0, 1, 0], [1, 0, 0], [1, 0, 0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1], [1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=False, alpha=1.0, gamma=0.0) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: focal_loss = sess.run(focal_loss) self.assertAllClose((((- math.log(0.5)) * 1.0) * 8), focal_loss) def testExpectedLossWithAlpha75AndZeroGamma(self): prediction_tensor = tf.constant([[[0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0]], [[0, 0, 0], [0, 0, 0], [0, 0, 0], [0, 0, 0]]], tf.float32) target_tensor = tf.constant([[[0, 1, 0], [1, 0, 0], [1, 0, 0], [0, 0, 1]], [[0, 0, 1], [0, 1, 0], [1, 0, 0], [1, 0, 0]]], tf.float32) weights = tf.constant([[1, 1, 1, 1], [1, 1, 1, 1]], tf.float32) focal_loss_op = losses.SigmoidFocalClassificationLoss(anchorwise_output=False, alpha=0.75, gamma=0.0) focal_loss = focal_loss_op(prediction_tensor, target_tensor, weights=weights) with self.test_session() as sess: focal_loss = sess.run(focal_loss) self.assertAllClose(((- math.log(0.5)) * ((0.75 * 8) + ((0.25 * 8) * 2))), focal_loss)
def detokenize(tokens, src_dict, idx2bpe): raw_inds = map(int, src_dict.string(tokens).split()) raw_chrs = ''.join([idx2bpe[raw_ind] for raw_ind in raw_inds]) raw_chrs = raw_chrs.replace('G', ' ') return raw_chrs
class EntropyMinTrainerHook(TrainerHook): def __init__(self, name: str, weight: float): super().__init__(name) self._weight = weight self._criterion = Entropy() def __call__(self): return _EntropyEpocherHook(name=self._hook_name, weight=self._weight, criterion=self._criterion)
def minimum_spanning_tree(weight_by_edge): group_by_node = {} mst_edges = set() for edge in sorted(weight_by_edge, key=weight_by_edge.__getitem__): (u, v) = edge if (group_by_node.setdefault(u, {u}) != group_by_node.setdefault(v, {v})): mst_edges.add(edge) group_by_node[u].update(group_by_node[v]) for node in group_by_node[v]: group_by_node[node].update(group_by_node[u]) return mst_edges
class Combinator(abc.ABC): def combine(self, data: List[np.ndarray]) -> np.ndarray: raise NotImplementedError()
def get_pi_trainable_variables(scope): return [v for v in get_trainable_variables(scope) if ('pi' in v.name[len(scope):].split('/'))]
def isstring(obj): try: return isinstance(obj, basestring) except NameError: return isinstance(obj, str)
def test_passed_sym_captured_as_dep_for_mutated_obj(): run_cell('\n class Foo:\n def __init__(self, x):\n self.x = x\n\n def mutate(foo, x):\n foo.x = x\n ') run_cell('foo = Foo(5)') run_cell('y = 7') run_cell('mutate(foo, y)') run_cell('logging.info(foo.x)') assert_not_detected() run_cell('y = 42') run_cell('logging.info(foo.x)') assert_detected('`foo.x` depends on old value of `y`')
def main(cfg, gpu, save_dir): start = time.time() best_mIoU = 0.0 best_epoch = 0 num_workers = 8 device = torch.device(cfg['DEVICE']) (train_cfg, eval_cfg) = (cfg['TRAIN'], cfg['EVAL']) (dataset_cfg, model_cfg) = (cfg['DATASET'], cfg['MODEL']) (loss_cfg, optim_cfg, sched_cfg) = (cfg['LOSS'], cfg['OPTIMIZER'], cfg['SCHEDULER']) (epochs, lr) = (train_cfg['EPOCHS'], optim_cfg['LR']) resume_path = cfg['MODEL']['RESUME'] gpus = int(os.environ['WORLD_SIZE']) traintransform = get_train_augmentation(train_cfg['IMAGE_SIZE'], seg_fill=dataset_cfg['IGNORE_LABEL']) valtransform = get_val_augmentation(eval_cfg['IMAGE_SIZE']) trainset = eval(dataset_cfg['NAME'])(dataset_cfg['ROOT'], 'train', traintransform, dataset_cfg['MODALS']) valset = eval(dataset_cfg['NAME'])(dataset_cfg['ROOT'], 'val', valtransform, dataset_cfg['MODALS']) class_names = trainset.CLASSES model = eval(model_cfg['NAME'])(model_cfg['BACKBONE'], trainset.n_classes, dataset_cfg['MODALS']) resume_checkpoint = None if os.path.isfile(resume_path): resume_checkpoint = torch.load(resume_path, map_location=torch.device('cpu')) msg = model.load_state_dict(resume_checkpoint['model_state_dict']) logger.info(msg) else: model.init_pretrained(model_cfg['PRETRAINED']) model = model.to(device) iters_per_epoch = ((len(trainset) // train_cfg['BATCH_SIZE']) // gpus) loss_fn = get_loss(loss_cfg['NAME'], trainset.ignore_label, None) start_epoch = 0 optimizer = get_optimizer(model, optim_cfg['NAME'], lr, optim_cfg['WEIGHT_DECAY']) scheduler = get_scheduler(sched_cfg['NAME'], optimizer, int(((epochs + 1) * iters_per_epoch)), sched_cfg['POWER'], (iters_per_epoch * sched_cfg['WARMUP']), sched_cfg['WARMUP_RATIO']) if train_cfg['DDP']: sampler = DistributedSampler(trainset, dist.get_world_size(), dist.get_rank(), shuffle=True) sampler_val = None model = DDP(model, device_ids=[gpu], output_device=0, find_unused_parameters=True) else: sampler = RandomSampler(trainset) sampler_val = None if resume_checkpoint: start_epoch = (resume_checkpoint['epoch'] - 1) optimizer.load_state_dict(resume_checkpoint['optimizer_state_dict']) scheduler.load_state_dict(resume_checkpoint['scheduler_state_dict']) loss = resume_checkpoint['loss'] best_mIoU = resume_checkpoint['best_miou'] trainloader = DataLoader(trainset, batch_size=train_cfg['BATCH_SIZE'], num_workers=num_workers, drop_last=True, pin_memory=False, sampler=sampler) valloader = DataLoader(valset, batch_size=eval_cfg['BATCH_SIZE'], num_workers=num_workers, pin_memory=False, sampler=sampler_val) scaler = GradScaler(enabled=train_cfg['AMP']) if ((train_cfg['DDP'] and (torch.distributed.get_rank() == 0)) or (not train_cfg['DDP'])): writer = SummaryWriter(str(save_dir)) logger.info(' model complexity ') cal_flops(model, dataset_cfg['MODALS'], logger) logger.info(' model structure ') logger.info(model) logger.info(' training config ') logger.info(cfg) for epoch in range(start_epoch, epochs): model.train() if train_cfg['DDP']: sampler.set_epoch(epoch) train_loss = 0.0 lr = scheduler.get_lr() lr = (sum(lr) / len(lr)) pbar = tqdm(enumerate(trainloader), total=iters_per_epoch, desc=f'Epoch: [{(epoch + 1)}/{epochs}] Iter: [{0}/{iters_per_epoch}] LR: {lr:.8f} Loss: {train_loss:.8f}') for (iter, (sample, lbl)) in pbar: optimizer.zero_grad(set_to_none=True) sample = [x.to(device) for x in sample] lbl = lbl.to(device) with autocast(enabled=train_cfg['AMP']): logits = model(sample) loss = loss_fn(logits, lbl) scaler.scale(loss).backward() scaler.step(optimizer) scaler.update() scheduler.step() torch.cuda.synchronize() lr = scheduler.get_lr() lr = (sum(lr) / len(lr)) if (lr <= 1e-08): lr = 1e-08 train_loss += loss.item() pbar.set_description(f'Epoch: [{(epoch + 1)}/{epochs}] Iter: [{(iter + 1)}/{iters_per_epoch}] LR: {lr:.8f} Loss: {(train_loss / (iter + 1)):.8f}') train_loss /= (iter + 1) if ((train_cfg['DDP'] and (torch.distributed.get_rank() == 0)) or (not train_cfg['DDP'])): writer.add_scalar('train/loss', train_loss, epoch) torch.cuda.empty_cache() if (((((epoch + 1) % train_cfg['EVAL_INTERVAL']) == 0) and ((epoch + 1) > train_cfg['EVAL_START'])) or ((epoch + 1) == epochs)): if ((train_cfg['DDP'] and (torch.distributed.get_rank() == 0)) or (not train_cfg['DDP'])): (acc, macc, _, _, ious, miou) = evaluate(model, valloader, device) writer.add_scalar('val/mIoU', miou, epoch) if (miou > best_mIoU): prev_best_ckp = (save_dir / f"{model_cfg['NAME']}_{model_cfg['BACKBONE']}_{dataset_cfg['NAME']}_epoch{best_epoch}_{best_mIoU}_checkpoint.pth") prev_best = (save_dir / f"{model_cfg['NAME']}_{model_cfg['BACKBONE']}_{dataset_cfg['NAME']}_epoch{best_epoch}_{best_mIoU}.pth") if os.path.isfile(prev_best): os.remove(prev_best) if os.path.isfile(prev_best_ckp): os.remove(prev_best_ckp) best_mIoU = miou best_epoch = (epoch + 1) cur_best_ckp = (save_dir / f"{model_cfg['NAME']}_{model_cfg['BACKBONE']}_{dataset_cfg['NAME']}_epoch{best_epoch}_{best_mIoU}_checkpoint.pth") cur_best = (save_dir / f"{model_cfg['NAME']}_{model_cfg['BACKBONE']}_{dataset_cfg['NAME']}_epoch{best_epoch}_{best_mIoU}.pth") torch.save((model.module.state_dict() if train_cfg['DDP'] else model.state_dict()), cur_best) torch.save({'epoch': best_epoch, 'model_state_dict': (model.module.state_dict() if train_cfg['DDP'] else model.state_dict()), 'optimizer_state_dict': optimizer.state_dict(), 'loss': train_loss, 'scheduler_state_dict': scheduler.state_dict(), 'best_miou': best_mIoU}, cur_best_ckp) logger.info(print_iou(epoch, ious, miou, acc, macc, class_names)) logger.info(f'Current epoch:{epoch} mIoU: {miou} Best mIoU: {best_mIoU}') if ((train_cfg['DDP'] and (torch.distributed.get_rank() == 0)) or (not train_cfg['DDP'])): writer.close() pbar.close() end = time.gmtime((time.time() - start)) table = [['Best mIoU', f'{best_mIoU:.2f}'], ['Total Training Time', time.strftime('%H:%M:%S', end)]] logger.info(tabulate(table, numalign='right'))
def wrap_time(func): def wrapped_func(*args, **kwargs): from time import time begin = time() func(*args, **kwargs) end = time() print(f"func {func.__name__}'s execution time: {(end - begin)}") return wrapped_func
def get_rank(): if (not dist.is_nccl_available()): return 0 if (not dist.is_initialized()): return 0 return dist.get_rank()
class UNetMidBlock2DCrossAttn(nn.Module): def __init__(self, in_channels: int, temb_channels: int, dropout: float=0.0, num_layers: int=1, transformer_layers_per_block: Union[(int, Tuple[int])]=1, resnet_eps: float=1e-06, resnet_time_scale_shift: str='default', resnet_act_fn: str='swish', resnet_groups: int=32, resnet_pre_norm: bool=True, num_attention_heads: int=1, output_scale_factor: float=1.0, cross_attention_dim: int=1280, dual_cross_attention: bool=False, use_linear_projection: bool=False, upcast_attention: bool=False, attention_type: str='default'): super().__init__() self.has_cross_attention = True self.num_attention_heads = num_attention_heads resnet_groups = (resnet_groups if (resnet_groups is not None) else min((in_channels // 4), 32)) if isinstance(transformer_layers_per_block, int): transformer_layers_per_block = ([transformer_layers_per_block] * num_layers) resnets = [ResnetBlock2D(in_channels=in_channels, out_channels=in_channels, temb_channels=temb_channels, eps=resnet_eps, groups=resnet_groups, dropout=dropout, time_embedding_norm=resnet_time_scale_shift, non_linearity=resnet_act_fn, output_scale_factor=output_scale_factor, pre_norm=resnet_pre_norm)] attentions = [] for i in range(num_layers): if (not dual_cross_attention): attentions.append(Transformer2DModel(num_attention_heads, (in_channels // num_attention_heads), in_channels=in_channels, num_layers=transformer_layers_per_block[i], cross_attention_dim=cross_attention_dim, norm_num_groups=resnet_groups, use_linear_projection=use_linear_projection, upcast_attention=upcast_attention, attention_type=attention_type)) else: attentions.append(DualTransformer2DModel(num_attention_heads, (in_channels // num_attention_heads), in_channels=in_channels, num_layers=1, cross_attention_dim=cross_attention_dim, norm_num_groups=resnet_groups)) resnets.append(ResnetBlock2D(in_channels=in_channels, out_channels=in_channels, temb_channels=temb_channels, eps=resnet_eps, groups=resnet_groups, dropout=dropout, time_embedding_norm=resnet_time_scale_shift, non_linearity=resnet_act_fn, output_scale_factor=output_scale_factor, pre_norm=resnet_pre_norm)) self.attentions = nn.ModuleList(attentions) self.resnets = nn.ModuleList(resnets) self.gradient_checkpointing = False def forward(self, hidden_states: torch.FloatTensor, temb: Optional[torch.FloatTensor]=None, encoder_hidden_states: Optional[torch.FloatTensor]=None, attention_mask: Optional[torch.FloatTensor]=None, cross_attention_kwargs: Optional[Dict[(str, Any)]]=None, encoder_attention_mask: Optional[torch.FloatTensor]=None) -> torch.FloatTensor: lora_scale = (cross_attention_kwargs.get('scale', 1.0) if (cross_attention_kwargs is not None) else 1.0) hidden_states = self.resnets[0](hidden_states, temb, scale=lora_scale) for (attn, resnet) in zip(self.attentions, self.resnets[1:]): if (self.training and self.gradient_checkpointing): def create_custom_forward(module, return_dict=None): def custom_forward(*inputs): if (return_dict is not None): return module(*inputs, return_dict=return_dict) else: return module(*inputs) return custom_forward ckpt_kwargs: Dict[(str, Any)] = ({'use_reentrant': False} if is_torch_version('>=', '1.11.0') else {}) hidden_states = attn(hidden_states, encoder_hidden_states=encoder_hidden_states, cross_attention_kwargs=cross_attention_kwargs, attention_mask=attention_mask, encoder_attention_mask=encoder_attention_mask, return_dict=False)[0] hidden_states = torch.utils.checkpoint.checkpoint(create_custom_forward(resnet), hidden_states, temb, **ckpt_kwargs) else: hidden_states = attn(hidden_states, encoder_hidden_states=encoder_hidden_states, cross_attention_kwargs=cross_attention_kwargs, attention_mask=attention_mask, encoder_attention_mask=encoder_attention_mask, return_dict=False)[0] hidden_states = resnet(hidden_states, temb, scale=lora_scale) return hidden_states
def main(): (args, *_) = cast(tuple[(Args, ...)], HfArgumentParser(Args).parse_args_into_dataclasses()) split = (f'train[:{args.max_considered_data}]' if (args.max_considered_data is not None) else 'train') assert (magicoder.utils.OPENAI_CLIENT is not None) dataset: Dataset = load_dataset(args.dataset_name, data_dir=args.data_dir, split=split, num_proc=magicoder.utils.N_CORES) random.seed(args.seed) dataset = dataset.map(function=map_dataset, fn_kwargs=dict(args=args), with_indices=True, batched=True, batch_size=args.chunk_size) dataset = dataset.shuffle(seed=args.seed) dataset = dataset.map((lambda _, index: {'index': index}), with_indices=True) start_index = args.seed_code_start_index end_index = min((start_index + args.max_new_data), len(dataset)) dataset = dataset.select(range(start_index, end_index)) prompt_template = Path('data/prompt.txt').read_text() timestamp = magicoder.utils.timestamp() data_fingerprint = args.fingerprint(prompt_template) if (args.continue_from is not None): assert (data_fingerprint in args.continue_from), 'Fingerprint mismatch' assert (f'{start_index}_{end_index}' in args.continue_from), 'Index mismatch' old_path = Path(args.continue_from) assert old_path.exists() old_data = magicoder.utils.read_jsonl(old_path) assert (len(old_data) > 0) last_index = old_data[(- 1)]['index'] n_skipped = ((last_index - start_index) + 1) print('Continuing from', old_path) f_out = old_path.open('a') else: tag = ('' if (args.tag == '') else f'-{args.tag}') path = Path(f'data{tag}-{data_fingerprint}-{start_index}_{end_index}-{timestamp}.jsonl') assert (not path.exists()) f_out = path.open('w') print('Saving to', path) n_skipped = 0 for (index, example) in enumerate(tqdm(dataset)): if (index < n_skipped): continue assert ((index + start_index) == example['index']) prompt = prompt_template.format(code=example['seed']) max_new_tokens = min(args.max_new_tokens, ((args.model_max_tokens - magicoder.utils.num_tokens_from_string(prompt, args.model)) - ERROR_MARGIN)) if (max_new_tokens <= 0): continue messages = [{'role': 'system', 'content': SYSTEM}, {'role': 'user', 'content': prompt}] openai_seed = (args.seed + example['index']) response = magicoder.utils.chat_completions_with_backoff(model=args.model, messages=messages, max_tokens=max_new_tokens, n=1, temperature=args.temperature, seed=openai_seed) print(openai_seed) choice = response.choices[0] if (choice.finish_reason != 'stop'): continue parsing_result = parse_problem_solution(choice.message.content) if (parsing_result is None): continue (problem, solution) = parsing_result if ((len(problem) == 0) or (len(solution) == 0)): continue fingerprint = response.system_fingerprint assert (fingerprint is not None) data = dict(raw_index=example['raw_index'], index=example['index'], seed=example['seed'], openai_fingerprint=fingerprint, problem=problem, solution=solution) print('[Problem Description]', problem, sep='\n', end='\n\n') print('[Solution]', solution, sep='\n') f_out.write((json.dumps(data) + '\n'))
class RecursiveListEnv(Environment): def __init__(self, length=10, encoding_dim=32): assert (length > 1), 'length must be a positive integer superior to 1' self.length = length self.start_pos = 0 self.end_pos = (length - 1) self.scratchpad_ints = np.zeros((length,)) self.p1_pos = 0 self.p2_pos = 0 self.encoding_dim = encoding_dim self.has_been_reset = False self.programs_library = {'PTR_1_LEFT': {'level': 0, 'recursive': False}, 'STOP': {'level': (- 1), 'recursive': False}, 'PTR_2_LEFT': {'level': 0, 'recursive': False}, 'PTR_1_RIGHT': {'level': 0, 'recursive': False}, 'PTR_2_RIGHT': {'level': 0, 'recursive': False}, 'SWAP': {'level': 0, 'recursive': False}, 'RSHIFT': {'level': 1, 'recursive': False}, 'LSHIFT': {'level': 1, 'recursive': False}, 'COMPSWAP': {'level': 1, 'recursive': False}, 'RESET': {'level': 2, 'recursive': True}, 'BUBBLE': {'level': 2, 'recursive': True}, 'BUBBLESORT': {'level': 3, 'recursive': True}} for (i, key) in enumerate(sorted(list(self.programs_library.keys()))): self.programs_library[key]['index'] = i self.prog_to_func = {'STOP': self._stop, 'PTR_1_LEFT': self._ptr_1_left, 'PTR_2_LEFT': self._ptr_2_left, 'PTR_1_RIGHT': self._ptr_1_right, 'PTR_2_RIGHT': self._ptr_2_right, 'SWAP': self._swap} self.prog_to_precondition = {'STOP': self._stop_precondition, 'RSHIFT': self._rshift_precondition, 'LSHIFT': self._lshift_precondition, 'COMPSWAP': self._compswap_precondition, 'RESET': self._reset_precondition, 'BUBBLE': self._bubble_precondition, 'BUBBLESORT': self._bubblesort_precondition, 'PTR_1_LEFT': self._ptr_1_left_precondition, 'PTR_2_LEFT': self._ptr_2_left_precondition, 'PTR_1_RIGHT': self._ptr_1_right_precondition, 'PTR_2_RIGHT': self._ptr_2_right_precondition, 'SWAP': self._swap_precondition} self.prog_to_postcondition = {'RSHIFT': self._rshift_postcondition, 'LSHIFT': self._lshift_postcondition, 'COMPSWAP': self._compswap_postcondition, 'RESET': self._reset_postcondition, 'BUBBLE': self._bubble_postcondition, 'BUBBLESORT': self._bubblesort_postcondition} super(RecursiveListEnv, self).__init__(self.programs_library, self.prog_to_func, self.prog_to_precondition, self.prog_to_postcondition) def _decr_length_right(self): assert self._decr_length_right_precondition(), 'precondition not verified' if (self.end_pos > self.start_pos): self.end_pos -= 1 def _decr_length_right_precondition(self): return ((self.end_pos > self.start_pos) and ((self.p1_pos < self.end_pos) and (self.p2_pos < self.end_pos))) def _decr_length_left(self): assert self._decr_length_left_precondition(), 'precondition not verified' if (self.start_pos < self.end_pos): self.start_pos += 1 def _decr_length_left_precondition(self): return ((self.start_pos < self.end_pos) and ((self.p1_pos > self.start_pos) and (self.p2_pos > self.start_pos))) def _incr_length_left(self): assert self._incr_length_left_precondition(), 'precondition not verified' if (self.start_pos > 0): self.start_pos -= 1 def _incr_length_left_precondition(self): return (self.start_pos > 0) def _incr_length_right(self): assert self._incr_length_right_precondition(), 'precondition not verified' if (self.end_pos < (self.length - 1)): self.end_pos += 1 def _incr_length_right_precondition(self): return (self.end_pos < (self.length - 1)) def _ptr_1_left(self): assert self._ptr_1_left_precondition(), 'precondition not verified' if (self.p1_pos > self.start_pos): self.p1_pos -= 1 def _ptr_1_left_precondition(self): return (self.p1_pos > self.start_pos) def _stop(self): assert self._stop_precondition(), 'precondition not verified' pass def _stop_precondition(self): return True def _ptr_2_left(self): assert self._ptr_2_left_precondition(), 'precondition not verified' if (self.p2_pos > self.start_pos): self.p2_pos -= 1 def _ptr_2_left_precondition(self): return (self.p2_pos > self.start_pos) def _ptr_1_right(self): assert self._ptr_1_right_precondition(), 'precondition not verified' if (self.p1_pos < self.end_pos): self.p1_pos += 1 def _ptr_1_right_precondition(self): return (self.p1_pos < self.end_pos) def _ptr_2_right(self): assert self._ptr_2_right_precondition(), 'precondition not verified' if (self.p2_pos < self.end_pos): self.p2_pos += 1 def _ptr_2_right_precondition(self): return (self.p2_pos < self.end_pos) def _swap(self): assert self._swap_precondition(), 'precondition not verified' self.scratchpad_ints[[self.p1_pos, self.p2_pos]] = self.scratchpad_ints[[self.p2_pos, self.p1_pos]] def _swap_precondition(self): return (self.p1_pos != self.p2_pos) def _compswap_precondition(self): list_length = ((self.end_pos - self.start_pos) + 1) bool = (list_length > 1) bool &= (self.p1_pos < self.end_pos) bool &= ((self.p2_pos == self.p1_pos) or (self.p2_pos == (self.p1_pos + 1))) return bool def _compswap_postcondition(self, init_state, state): (new_scratchpad_ints, new_p1_pos, new_p2_pos, new_start_pos, new_end_pos) = init_state new_scratchpad_ints = np.copy(new_scratchpad_ints) if ((new_p1_pos == new_p2_pos) and (new_p2_pos < new_end_pos)): new_p2_pos += 1 idx_left = min(new_p1_pos, new_p2_pos) idx_right = max(new_p1_pos, new_p2_pos) if (new_scratchpad_ints[idx_left] > new_scratchpad_ints[idx_right]): new_scratchpad_ints[[idx_left, idx_right]] = new_scratchpad_ints[[idx_right, idx_left]] new_state = (new_scratchpad_ints, new_p1_pos, new_p2_pos, new_start_pos, new_end_pos) return self.compare_state(state, new_state) def _lshift_postcondition(self, init_state, state): (init_scratchpad_ints, init_p1_pos, init_p2_pos, init_start_pos, init_end_pos) = init_state (scratchpad_ints, p1_pos, p2_pos, start_pos, end_pos) = state bool = np.array_equal(init_scratchpad_ints, scratchpad_ints) bool &= (init_start_pos == start_pos) bool &= (init_end_pos == end_pos) if (init_p1_pos > init_start_pos): bool &= (p1_pos == (init_p1_pos - 1)) else: bool &= (p1_pos == init_p1_pos) if (init_p2_pos > init_start_pos): bool &= (p2_pos == (init_p2_pos - 1)) else: bool &= (p2_pos == init_p2_pos) return bool def _rshift_postcondition(self, init_state, state): (init_scratchpad_ints, init_p1_pos, init_p2_pos, init_start_pos, init_end_pos) = init_state (scratchpad_ints, p1_pos, p2_pos, start_pos, end_pos) = state bool = np.array_equal(init_scratchpad_ints, scratchpad_ints) bool &= (init_start_pos == start_pos) bool &= (init_end_pos == end_pos) if (init_p1_pos < init_end_pos): bool &= (p1_pos == (init_p1_pos + 1)) else: bool &= (p1_pos == init_p1_pos) if (init_p2_pos < init_end_pos): bool &= (p2_pos == (init_p2_pos + 1)) else: bool &= (p2_pos == init_p2_pos) return bool def _reset_postcondition(self, init_state, state): (init_scratchpad_ints, init_p1_pos, init_p2_pos, init_start_pos, init_end_pos) = init_state (scratchpad_ints, p1_pos, p2_pos, start_pos, end_pos) = state bool = np.array_equal(init_scratchpad_ints, scratchpad_ints) bool &= (init_start_pos == start_pos) bool &= (init_end_pos == end_pos) bool &= ((p1_pos == start_pos) and (p2_pos == start_pos)) return bool def _bubblesort_postcondition(self, init_state, state): (init_scratchpad_ints, init_p1_pos, init_p2_pos, init_start_pos, init_end_pos) = init_state (scratchpad_ints, p1_pos, p2_pos, start_pos, end_pos) = state bool = (init_start_pos == start_pos) bool &= (init_end_pos == end_pos) bool &= np.all((scratchpad_ints[:end_pos] <= scratchpad_ints[(start_pos + 1):(end_pos + 1)])) return bool def _bubble_postcondition(self, init_state, state): (new_scratchpad_ints, new_p1_pos, new_p2_pos, new_start_pos, new_end_pos) = init_state new_scratchpad_ints = np.copy(new_scratchpad_ints) for idx in range(new_start_pos, new_end_pos): if (new_scratchpad_ints[(idx + 1)] < new_scratchpad_ints[idx]): new_scratchpad_ints[[idx, (idx + 1)]] = new_scratchpad_ints[[(idx + 1), idx]] new_p1_pos = new_end_pos new_p2_pos = new_end_pos new_state = (new_scratchpad_ints, new_p1_pos, new_p2_pos, new_start_pos, new_end_pos) return self.compare_state(state, new_state) def _lshift_precondition(self): return ((self.p1_pos > self.start_pos) or (self.p2_pos > self.start_pos)) def _rshift_precondition(self): return ((self.p1_pos < self.end_pos) or (self.p2_pos < self.end_pos)) def _bubble_precondition(self): bubble_index = self.programs_library['BUBBLE']['index'] if (self.current_task_index != bubble_index): bool = (self.p1_pos == self.start_pos) bool &= ((self.p2_pos == self.start_pos) or (self.p2_pos == (self.start_pos + 1))) else: bool = (self.p1_pos == (self.start_pos + 1)) bool &= ((self.p2_pos == self.start_pos) or (self.p2_pos == (self.start_pos + 1))) bool &= self._decr_length_left_precondition() return bool def _reset_precondition(self): bool = ((self.p1_pos > self.start_pos) or (self.p2_pos > self.start_pos)) reset_index = self.programs_library['RESET']['index'] if (self.current_task_index == reset_index): bool &= self._decr_length_right_precondition() return bool def _bubblesort_precondition(self): bool = (self.p1_pos == self.start_pos) bool &= (self.p2_pos == self.start_pos) bubblesort_index = self.programs_library['BUBBLESORT']['index'] if (self.current_task_index == bubblesort_index): bool &= self._decr_length_right_precondition() return bool def _one_hot_encode(self, digit, basis=10): encoding = np.zeros(basis) encoding[digit] = 1 return encoding def _one_hot_decode(self, one_encoding): return np.argmax(one_encoding) def reset_env(self): assert (self.length > 1), 'list length must be greater than 1' self.start_pos = 0 self.end_pos = (self.length - 1) self.scratchpad_ints = np.random.randint(10, size=self.length) current_task_name = self.get_program_from_index(self.current_task_index) if ((current_task_name == 'BUBBLE') or (current_task_name == 'BUBBLESORT')): init_pointers_pos1 = 0 init_pointers_pos2 = 0 elif (current_task_name == 'RESET'): while True: init_pointers_pos1 = int(np.random.randint(0, self.length)) init_pointers_pos2 = int(np.random.randint(0, self.length)) if (not ((init_pointers_pos1 == 0) and (init_pointers_pos2 == 0))): break elif (current_task_name == 'LSHIFT'): while True: init_pointers_pos1 = int(np.random.randint(0, self.length)) init_pointers_pos2 = int(np.random.randint(0, self.length)) if (not ((init_pointers_pos1 == 0) and (init_pointers_pos2 == 0))): break elif (current_task_name == 'RSHIFT'): while True: init_pointers_pos1 = int(np.random.randint(0, self.length)) init_pointers_pos2 = int(np.random.randint(0, self.length)) if (not ((init_pointers_pos1 == (self.length - 1)) and (init_pointers_pos2 == (self.length - 1)))): break elif (current_task_name == 'COMPSWAP'): init_pointers_pos1 = int(np.random.randint(0, (self.length - 1))) init_pointers_pos2 = int(np.random.choice([init_pointers_pos1, (init_pointers_pos1 + 1)])) else: raise NotImplementedError('Unable to reset env for this program...') self.p1_pos = init_pointers_pos1 self.p2_pos = init_pointers_pos2 self.has_been_reset = True def get_observation(self): assert self.has_been_reset, 'Need to reset the environment before getting observations' p1_val = self.scratchpad_ints[self.p1_pos] p2_val = self.scratchpad_ints[self.p2_pos] is_sorted = int(self._is_sorted()) pointers_same_pos = int((self.p1_pos == self.p2_pos)) pt_1_left = int((self.p1_pos == self.start_pos)) pt_2_left = int((self.p2_pos == self.start_pos)) pt_1_right = int((self.p1_pos == self.end_pos)) pt_2_right = int((self.p2_pos == self.end_pos)) p1p2 = np.eye(10)[[p1_val, p2_val]].reshape((- 1)) bools = np.array([pt_1_left, pt_1_right, pt_2_left, pt_2_right, pointers_same_pos, is_sorted]) return np.concatenate((p1p2, bools), axis=0) def get_observation_dim(self): return ((2 * 10) + 6) def get_state(self): assert self.has_been_reset, 'Need to reset the environment before getting states' return (np.copy(self.scratchpad_ints), self.p1_pos, self.p2_pos, self.start_pos, self.end_pos) def reset_to_state(self, state): self.scratchpad_ints = state[0].copy() self.p1_pos = state[1] self.p2_pos = state[2] self.start_pos = state[3] self.end_pos = state[4] def get_state_str(self, state): scratchpad = state[0].copy() p1_pos = state[1] p2_pos = state[2] start_pos = state[3] end_pos = state[4] scratchpad = scratchpad[start_pos:(end_pos + 1)] str = 'list: {}, p1 : {}, p2 : {}, start_pos: {}, end_pos: {}'.format(scratchpad, p1_pos, p2_pos, start_pos, end_pos) return str def compare_state(self, state1, state2): bool = np.array_equal(state1[0], state2[0]) bool &= (state1[1] == state2[1]) bool &= (state1[2] == state2[2]) bool &= (state1[3] == state2[3]) bool &= (state1[4] == state2[4]) return bool def _is_sorted(self): arr = self.scratchpad_ints return np.all((arr[:(self.length - 1)] <= arr[1:self.length])) def start_task(self, task_index): if (self.tasks_list.count(task_index) > 0): task = self.get_program_from_index(task_index) if ((task == 'RESET') or (task == 'BUBBLESORT')): self._decr_length_right() if (task == 'BUBBLE'): self._decr_length_left() return super(RecursiveListEnv, self).start_task(task_index) def end_task(self): current_task = self.get_program_from_index(self.current_task_index) if ((current_task == 'RESET') or (current_task == 'BUBBLESORT')): if (self.tasks_list.count(self.current_task_index) > 1): self._incr_length_right() if (current_task == 'BUBBLE'): if (self.tasks_list.count(self.current_task_index) > 1): self._incr_length_left() super(RecursiveListEnv, self).end_task()
_registry class TensorFlowAdaptor(Adaptor): unify_op_type_mapping = {'Conv2D': 'conv2d', 'Conv3D': 'conv3d', 'DepthwiseConv2dNative': 'conv2d', 'FusedBatchNormV3': 'batchnorm', '_MklFusedInstanceNorm': 'instancenorm', 'MaxPool': 'pooling', 'MaxPool3D': 'pooling', 'AvgPool': 'pooling', 'ConcatV2': 'concat', 'MatMul': 'matmul', 'BatchMatMul': 'matmul', 'BatchMatMulV2': 'matmul', 'Pad': 'pad', 'Conv2DBackpropInput': 'deconv2d', 'Conv3DBackpropInputV2': 'deconv3d'} def __init__(self, framework_specific_info): super().__init__(framework_specific_info) os.environ['CUDA_DEVICE_ORDER'] = 'PCI_BUS_ID' os.environ['CUDA_VISIBLE_DEVICES'] = '-1' self.quantize_config = {'op_wise_config': {}} self.framework_specific_info = framework_specific_info self.approach = deep_get(self.framework_specific_info, 'approach', False) self.device = self.framework_specific_info['device'] self.work_dir = os.path.abspath(self.framework_specific_info['workspace_path']) self.recipes = deep_get(self.framework_specific_info, 'recipes', {}) self.performance_only = deep_get(self.framework_specific_info, 'performance_only', False) self.use_bf16 = deep_get(self.framework_specific_info, 'use_bf16', False) self.backend = self.framework_specific_info['backend'] self.format = self.framework_specific_info['format'] os.makedirs(self.work_dir, exist_ok=True) self.model = None self.pre_optimized_model = None self.pre_optimizer_handle = None self.bf16_ops = [] self.fp32_ops = [] self.smooth_quant_mul_ops = [] self.dump_times = 0 cfg_yaml_name = '{}.yaml'.format(self.__class__.__name__[:(- len('Adaptor'))].lower()) self.itex_mode = ((self.backend == 'itex') or (cfg_yaml_name == 'tensorflow_itex.yaml')) if self.itex_mode: self._check_itex() self.query_handler = TensorflowQuery(local_config_file=os.path.join(os.path.dirname(__file__), cfg_yaml_name), performance_only=self.performance_only, itex_mode=self.itex_mode) import tensorflow as tf from pkg_resources import parse_version self.new_api = (tf.version.VERSION in spr_base_verions) self.qdq_enabled = (self.itex_mode or (self.format == 'QDQ') or self.new_api) self.op_wise_sequences = self.query_handler.get_eightbit_patterns(self.qdq_enabled) self.fp32_results = [] self.fp32_preds_as_label = False self.benchmark = (GLOBAL_STATE.STATE == MODE.BENCHMARK) self.callbacks = [] self.optype_statistics = None self._last_dequantize_ops = None self.smooth_quant_model = None def _check_itex(self): try: import intel_extension_for_tensorflow except: raise ImportError('The Intel Extension for TensorFlow is not installed. Please install it to run models on ITEX backend') def _log_histogram(self, writer, tag, values, step=0, bins=1000): import tensorflow as tf values = np.array(values) with writer.as_default(): tf.summary.histogram(tag, values, step) writer.flush() def _pre_hook_for_hvd(self, dataloader=None): import horovod.tensorflow as hvd self.hvd = hvd self.hvd.init() _elapsed_time(customized_msg='Model training') def train(self, model, dataloader, optimizer_tuple, criterion_tuple, hooks, postprocess, **kwargs): import tensorflow as tf from neural_compressor.model.tensorflow_model import get_model_type tf.random.set_seed(1) self.model_type = get_model_type(model._model) optimizer = optimizer_tuple[0](**optimizer_tuple[1]) criterion = criterion_tuple[0](**criterion_tuple[1]) start_epochs = kwargs['kwargs'].get('start_epoch', None) end_epochs = kwargs['kwargs'].get('end_epoch', None) epochs = kwargs['kwargs'].get('epoch', None) iters = kwargs['kwargs'].get('iteration', None) callbacks = kwargs['kwargs'].get('callbacks', None) execution_mode = kwargs['kwargs'].get('execution_mode', None) distributed = getattr(dataloader, 'distributed', False) from neural_compressor.compression.distillation.criterions import TensorflowKnowledgeDistillationLoss if isinstance(criterion, TensorflowKnowledgeDistillationLoss): input_model = model._model else: input_model = tf.keras.models.load_model(model._model) hooks = callbacks['tf_pruning'](model, input_model, hooks) hooks['on_train_begin']() train_loss_results = [] if distributed: try: len_dataloader = len(dataloader) except: logger.info('The length of the distributed training dataloader is unknown.When the iteration of training dataloader in each process is inconsistent, an error may occur.') else: list_len_dataloader = self.hvd.allgather_object(len_dataloader) if (self.hvd.rank() == 0): for i in range((len(list_len_dataloader) - 1)): if (list_len_dataloader[i] != list_len_dataloader[(i + 1)]): raise AttributeError("The training dataloader's iteration isdifferent between processes, please reset dataloader's batch_size.") def training_step(x, y, first_batch): with tf.GradientTape() as tape: tape.watch(input_model.trainable_variables) y_ = input_model(x, training=True) loss_value = criterion(y, y_) loss_value = hooks['on_after_compute_loss'](x, y_, loss_value) tape = (self.hvd.DistributedGradientTape(tape) if distributed else tape) grads = tape.gradient(loss_value, input_model.trainable_variables) optimizer.apply_gradients(zip(grads, input_model.trainable_variables)) if (distributed and first_batch): self.hvd.broadcast_variables(input_model.variables, root_rank=0) self.hvd.broadcast_variables(optimizer.variables(), root_rank=0) return loss_value training_step = (training_step if (execution_mode == 'eager') else tf.function(training_step)) if ((start_epochs is not None) and (end_epochs is not None)): epochs = (end_epochs - start_epochs) for epoch in range(epochs): cnt = 0 epoch_loss_avg = tf.keras.metrics.Mean() hooks['on_epoch_begin'](epoch) for (iter, data) in enumerate(dataloader): (x, y) = (postprocess(data) if (postprocess is not None) else data) hooks['on_step_begin'](iter) cnt += 1 loss_value = training_step(x, y, (iter == 0)) epoch_loss_avg.update_state(loss_value) hooks['on_step_end']() if ((iters is not None) and (cnt >= iters)): break model._sess = None hooks['on_epoch_end']() train_loss_results.append(epoch_loss_avg.result()) if distributed: logger.info('Epoch-{:03d} training on rank {!s} have been done.'.format((epoch + 1), self.hvd.allgather_object(self.hvd.rank()))) logger.info('Epoch {:03d}: Loss: {:.3f}'.format((epoch + 1), epoch_loss_avg.result())) hooks['on_train_end']() model._sess = None if (not isinstance(criterion, TensorflowKnowledgeDistillationLoss)): if distributed: if (self.hvd.rank() == 0): input_model.save(model._model) rank_list = self.hvd.allgather_object(self.hvd.rank()) logger.info(f"rank 0 has saved the pruned model to '{model._model}',all ranks {rank_list} ready.") else: input_model.save(model._model) else: input_model.save('distillation_model') _elapsed_time(customized_msg='Model inference') def evaluate(self, model, dataloader, postprocess=None, metrics=None, measurer=None, iteration=(- 1), tensorboard=False, fp32_baseline=False): import tensorflow as tf from .tf_utils.util import iterator_sess_run outputs = model.output_tensor_names if getattr(dataloader, 'distributed', False): import horovod.tensorflow as hvd hvd.init() for metric in metrics: metric.hvd = hvd try: len_dataloader = len(dataloader) except: logger.info('The length of the distributed evaluation dataloader is unknown.When the iteration of evaluation dataloader in each process is inconsistent, an error may occur.') else: list_len_dataloader = hvd.allgather_object(len_dataloader) if (hvd.rank() == 0): for i in range((len(list_len_dataloader) - 1)): if (list_len_dataloader[i] != list_len_dataloader[(i + 1)]): raise AttributeError("The evaluation dataloader's iteration isdifferent between processes, please reset dataloader's batch_size.") logger.info("Rank {!s} dataloaders' data distribution balance check for evaluation have been finished.".format(hvd.allgather_object(hvd.rank()))) if tensorboard: from tensorflow.python.framework import tensor_util from .tf_utils.graph_util import GraphAnalyzer output_postfix = '_fp32.output' inspect_node_types = ['Conv2D', 'DepthwiseConv2dNative', 'MaxPool', 'AvgPool', 'ConcatV2', 'MatMul', 'FusedBatchNormV3', 'FusedBatchNorm', 'BiasAdd', '_MklFusedInstanceNorm', 'Relu', 'Relu6', 'Dequantize'] fp32_inspect_node_name = [] int8_inspect_node_name = [] q_node_scale = {} if (self.dump_times == 0): temp_dir = './runs/eval/baseline' else: temp_dir = ('./runs/eval/tune_' + str(self.dump_times)) if os.path.isdir(temp_dir): import shutil shutil.rmtree(temp_dir, ignore_errors=True) writer = tf.summary.create_file_writer(temp_dir) with writer.as_default(): tf.summary.graph(model.graph) cur_graph = GraphAnalyzer() cur_graph.graph = model.graph_def cur_graph.parse_graph() graph_info = cur_graph.node_name_details for node in model.graph_def.node: if (node.op in inspect_node_types): fp32_inspect_node_name.append(node.name) elif (node.op.find('Requantize') != (- 1)): out_min = (- 2) out_max = (- 1) if (node.op.find('Sum') != (- 1)): out_min = (- 5) out_max = (- 4) q_out_min = graph_info[node.input[out_min]].node.attr['value'].tensor.float_val[0] q_out_max = graph_info[node.input[out_max]].node.attr['value'].tensor.float_val[0] q_node_scale[node.name] = (node.op, q_out_min, q_out_max) int8_inspect_node_name.append(node.name) if ((node.op == 'Const') and (graph_info[graph_info[node.name].outputs[0]].node.op in ['Conv2D', 'DepthwiseConv2dNative', 'MatMul', 'FusedBatchNormV3', '_MklFusedInstanceNorm', 'BiasAdd'])): const_value = tensor_util.MakeNdarray(node.attr.get('value').tensor).astype(np.float32) self._log_histogram(writer, node.name, const_value) outputs.extend(fp32_inspect_node_name) if (len(int8_inspect_node_name) > 0): output_postfix = '_int8.output' outputs.extend(int8_inspect_node_name) if metrics: for metric in metrics: metric.reset() self.fp32_preds_as_label = any([(hasattr(metric, 'compare_label') and (not metric.compare_label)) for metric in metrics]) origin_output_tensor_names = model.output_tensor_names model.output_tensor_names = outputs input_tensor = model.input_tensor output_tensor = (model.output_tensor if (len(model.output_tensor) > 1) else model.output_tensor[0]) logger.info('Start to evaluate the TensorFlow model.') def eval_func(dataloader): results = [] for (idx, (inputs, labels)) in enumerate(dataloader): if (len(input_tensor) == 1): feed_dict = {} if (isinstance(inputs, dict) or isinstance(inputs, OrderedDict) or isinstance(inputs, UserDict)): for name in inputs: for tensor in input_tensor: pos = tensor.name.rfind(':') t_name = (tensor.name if (pos < 0) else tensor.name[:pos]) if (name == t_name): feed_dict[tensor] = inputs[name] break else: feed_dict = {input_tensor[0]: inputs} else: assert (len(input_tensor) == len(inputs)), 'inputs len must equal with input_tensor' feed_dict = {} if (isinstance(inputs, dict) or isinstance(inputs, OrderedDict) or isinstance(inputs, UserDict)): for name in inputs: for tensor in input_tensor: pos = tensor.name.rfind(':') t_name = (tensor.name if (pos < 0) else tensor.name[:pos]) if (name == t_name): feed_dict[tensor] = inputs[name] break else: feed_dict = dict(zip(input_tensor, inputs)) if model.iter_op: predictions = iterator_sess_run(model.sess, model.iter_op, feed_dict, output_tensor, iteration, measurer) elif (measurer is not None): measurer.start() predictions = model.sess.run(output_tensor, feed_dict) measurer.end() else: predictions = model.sess.run(output_tensor, feed_dict) if self.fp32_preds_as_label: (self.fp32_results.append(predictions) if fp32_baseline else results.append(predictions)) if ((idx == 0) and tensorboard): for (index, node_name) in enumerate(outputs): tensor = predictions[index] if (node_name in int8_inspect_node_name): tensor = Dequantize(predictions[index], q_node_scale[node_name]) self._log_histogram(writer, (node_name + output_postfix), tensor.astype(np.float32), idx) writer.close() if isinstance(predictions, list): if (len(origin_output_tensor_names) == 1): predictions = predictions[0] elif (len(origin_output_tensor_names) > 1): predictions = predictions[:len(origin_output_tensor_names)] if (postprocess is not None): (predictions, labels) = postprocess((predictions, labels)) if metrics: for metric in metrics: if ((not hasattr(metric, 'compare_label')) or (hasattr(metric, 'compare_label') and metric.compare_label)): metric.update(predictions, labels) if ((idx + 1) == iteration): break return results if (isinstance(dataloader, BaseDataLoader) and (not self.benchmark)): try: results = eval_func(dataloader) except Exception: logger.warning('Fail to forward with batch size={}, set to {} now.'.format(dataloader.batch_size, 1)) dataloader.batch(1) results = eval_func(dataloader) else: results = eval_func(dataloader) if self.fp32_preds_as_label: from .tf_utils.util import collate_tf_preds if fp32_baseline: results = collate_tf_preds(self.fp32_results) reference = results else: reference = collate_tf_preds(self.fp32_results) results = collate_tf_preds(results) for metric in metrics: if (hasattr(metric, 'compare_label') and (not metric.compare_label)): metric.update(results, reference) acc = (0 if (metrics is None) else [metric.result() for metric in metrics]) if tensorboard: new_dir = ((temp_dir + '_acc_') + str(acc)) writer.close() if os.path.isdir(new_dir): import shutil shutil.rmtree(new_dir, ignore_errors=True) os.rename(temp_dir, new_dir) self.dump_times += 1 model.output_tensor_names = origin_output_tensor_names return (acc if ((not isinstance(acc, list)) or (len(acc) > 1)) else acc[0]) def _tuning_cfg_to_fw(self, tuning_cfg): self.quantize_config['calib_iteration'] = tuning_cfg['calib_iteration'] self.quantize_config['device'] = self.device self.quantize_config['advance'] = deep_get(tuning_cfg, 'advance') fp32_ops = [] bf16_ops = [] dispatched_op_names = [j[0] for j in tuning_cfg['op']] invalid_op_names = [i for i in self.quantize_config['op_wise_config'] if (i not in dispatched_op_names)] for op_name in invalid_op_names: self.quantize_config['op_wise_config'].pop(op_name) for each_op_info in tuning_cfg['op']: op_name = each_op_info[0] if (tuning_cfg['op'][each_op_info]['activation']['dtype'] in ['fp32', 'bf16']): if (op_name in self.quantize_config['op_wise_config']): self.quantize_config['op_wise_config'].pop(op_name) if (tuning_cfg['op'][each_op_info]['activation']['dtype'] == 'fp32'): fp32_ops.append(op_name) if (tuning_cfg['op'][each_op_info]['activation']['dtype'] == 'bf16'): bf16_ops.append(op_name) continue is_perchannel = False bit = None if ('weight' in tuning_cfg['op'][each_op_info]): is_perchannel = (tuning_cfg['op'][each_op_info]['weight']['granularity'] == 'per_channel') weight_bit = (bit if bit else 7.0) algorithm = tuning_cfg['op'][each_op_info]['activation']['algorithm'] is_asymmetric = False if ('activation' in tuning_cfg['op'][each_op_info]): is_asymmetric = (tuning_cfg['op'][each_op_info]['activation']['scheme'] == 'asym') self.quantize_config['op_wise_config'][op_name] = (is_perchannel, algorithm, is_asymmetric, weight_bit) self.fp32_ops = fp32_ops self.bf16_ops = bf16_ops _elapsed_time('Pass quantize model') def quantize(self, tune_cfg, model, data_loader, q_func=None): assert (self.approach != 'post_training_dynamic_quant'), 'Dynamic quantization is not supported on TensorFlow framework now!' if (self.approach == 'quant_aware_training'): assert (q_func is not None), 'quantization aware training mode is not configured correctly' from neural_compressor.model import Model qat_model = q_func(model) return self.convert(Model(qat_model), 'QAT', 'default') assert (q_func is None), 'post-training quantization mode is not support calibration function for Tensorflow!' self._tuning_cfg_to_fw(tune_cfg) self.bf16_ops.extend(self.smooth_quant_mul_ops) logger.debug('Dump quantization configurations:') logger.debug(self.quantize_config) from .tf_utils.graph_converter import GraphConverter calib_sampling_size = tune_cfg.get('calib_sampling_size', 1) if isinstance(data_loader, BaseDataLoader): batch_size = data_loader.batch_size try: for i in range(batch_size): if ((calib_sampling_size % (batch_size - i)) == 0): calib_batch_size = (batch_size - i) if (i != 0): logger.warning(('Reset `calibration.dataloader.batch_size` field to {}'.format(calib_batch_size) + ' to make sure the sampling_size is divisible exactly by batch size')) break tmp_iterations = int(math.ceil((calib_sampling_size / calib_batch_size))) data_loader.batch(calib_batch_size) self.quantize_config['calib_iteration'] = tmp_iterations converted_model = GraphConverter(model, qt_config=self.quantize_config, recipes=self.recipes, int8_sequences=self.op_wise_sequences, fp32_ops=self.fp32_ops, bf16_ops=self.bf16_ops, data_loader=data_loader, calib_func=q_func, qdq_enabled=self.qdq_enabled, new_api=self.new_api, performance_only=self.performance_only, use_bf16=self.use_bf16).convert() except Exception: from .tf_utils.util import get_model_input_shape batch_size = get_model_input_shape(model) logger.warning('Fail to forward with batch size={}, set to {} now.'.format(data_loader.batch_size, batch_size)) data_loader.batch(batch_size) self.quantize_config['calib_iteration'] = calib_sampling_size converted_model = GraphConverter(model, qt_config=self.quantize_config, recipes=self.recipes, int8_sequences=self.op_wise_sequences, fp32_ops=self.fp32_ops, bf16_ops=self.bf16_ops, data_loader=data_loader, calib_func=q_func, qdq_enabled=self.qdq_enabled, new_api=self.new_api, performance_only=self.performance_only, use_bf16=self.use_bf16).convert() else: if (hasattr(data_loader, 'batch_size') and ((calib_sampling_size % data_loader.batch_size) != 0)): iter = self.quantize_config['calib_iteration'] logger.warning("Please note that calibration sampling size {} isn't divisible exactly by batch size {}. So the real sampling size is {}.".format(calib_sampling_size, data_loader.batch_size, (data_loader.batch_size * iter))) converted_model = GraphConverter(model, qt_config=self.quantize_config, recipes=self.recipes, int8_sequences=self.op_wise_sequences, fp32_ops=self.fp32_ops, bf16_ops=self.bf16_ops, data_loader=data_loader, calib_func=q_func, qdq_enabled=self.qdq_enabled, new_api=self.new_api, performance_only=self.performance_only, use_bf16=self.use_bf16).convert() converted_model.q_config.update({'framework_specific_info': self.framework_specific_info}) self._dump_model_op_stats(converted_model.graph_def) return converted_model def _dump_model_op_stats(self, model_graphdef): fp32_op_list_uint8 = copy.deepcopy(self.query_handler.get_op_types_by_precision(precision='uint8')) fp32_op_list_int8 = copy.deepcopy(self.query_handler.get_op_types_by_precision(precision='int8')) fp32_op_list = list(set(fp32_op_list_uint8).union(set(fp32_op_list_int8))) int8_op_prefix_list = ['QuantizedConv2D', '_FusedQuantizedConv3D', 'QuantizedDepthwise', 'QuantizedMaxPool', 'QuantizedAvgPool', 'QuantizedConcatV2', 'QuantizedMatMul', '_QuantizedFusedBatchNorm', '_QuantizedMatMul', '_QuantizedBatchMatMul', '_QuantizedFusedInstanceNorm', '_FusedQuantizedDeconv2D', '_FusedQuantizedDeconv3D'] from tensorflow.python.framework import dtypes res = {} for op_type in fp32_op_list: res[op_type] = {'INT8': 0, 'BF16': 0, 'FP32': 0} res['QuantizeV2'] = {'INT8': 0, 'BF16': 0, 'FP32': 0} res['Dequantize'] = {'INT8': 0, 'BF16': 0, 'FP32': 0} res['Cast'] = {'INT8': 0, 'BF16': 0, 'FP32': 0} fp32_op_list.extend(['QuantizeV2', 'Dequantize', 'Cast']) for i in model_graphdef.node: if (i.op == 'Const'): continue possible_int8_res = [name for name in int8_op_prefix_list if (i.op.find(name) != (- 1))] if any(possible_int8_res): origin_op_type = possible_int8_res[0].split('Quantized')[(- 1)] if (origin_op_type == 'FusedBatchNorm'): origin_op_type = 'FusedBatchNormV3' if (origin_op_type == 'FusedInstanceNorm'): origin_op_type = '_MklFusedInstanceNorm' if (origin_op_type == 'Depthwise'): origin_op_type = 'DepthwiseConv2dNative' if (origin_op_type == 'BatchMatMul'): origin_op_type = 'BatchMatMulV2' if (origin_op_type == 'FusedBatchMatMulV2'): origin_op_type = '_MklFusedBatchMatMulV2' if (origin_op_type == 'Deconv2D'): origin_op_type = 'Conv2DBackpropInput' if (origin_op_type == 'Deconv3D'): origin_op_type = 'Conv3DBackpropInputV2' res[origin_op_type]['INT8'] += 1 if (i.op in fp32_op_list): if (('T' not in i.attr) and (i.op != 'Cast')): continue if (i.op == 'Cast'): if (i.attr['DstT'].type == dtypes.bfloat16): res[i.op]['BF16'] += 1 elif (i.attr['DstT'].type == dtypes.float32): res[i.op]['FP32'] += 1 elif (i.attr['T'].type == dtypes.bfloat16): res[i.op]['BF16'] += 1 elif (i.attr['T'].type in (dtypes.quint8, dtypes.qint8)): res[i.op]['INT8'] += 1 else: res[i.op]['FP32'] += 1 field_names = ['Op Type', 'Total', 'INT8', 'BF16', 'FP32'] output_data = [[op_type, sum(res[op_type].values()), res[op_type]['INT8'], res[op_type]['BF16'], res[op_type]['FP32']] for op_type in fp32_op_list] Statistics(output_data, header='Mixed Precision Statistics', field_names=field_names).print_stat() self.optype_statistics = (field_names, output_data) def _query_bf16_ops(self, matched_nodes): self.bf16_op_details = OrderedDict() valid_precision = self.query_handler.get_mixed_precision_combination() if ((('bf16' in valid_precision) and CpuInfo().bf16) or (os.getenv('FORCE_BF16') == '1')): for details in matched_nodes: node_op = details[(- 1)][0] node_name = details[0] self.bf16_op_details[(node_name, node_op)] = [{'weight': {'dtype': ['bf16']}, 'activation': {'dtype': ['bf16']}}, {'weight': {'dtype': 'fp32'}, 'activation': {'dtype': 'fp32'}}] def _query_quantizable_ops(self, matched_nodes): bf16_common_config = {'weight': {'dtype': 'bf16'}, 'activation': {'dtype': 'bf16'}} fp32_common_config = {'weight': {'dtype': 'fp32'}, 'activation': {'dtype': 'fp32'}} uint8_type = self.query_handler.get_op_types_by_precision(precision='uint8') int8_type = self.query_handler.get_op_types_by_precision(precision='int8') bf16_type = self.query_handler.get_op_types_by_precision(precision='bf16') tf_quantizable_op_type = list(set(uint8_type).union(set(int8_type))) valid_precision = self.query_handler.get_mixed_precision_combination() op_capability = self.query_handler.get_quantization_capability() conv_config = copy.deepcopy(op_capability['Conv2D']) conv3d_config = (copy.deepcopy(op_capability['Conv3D']) if ('Conv3D' in op_capability) else None) matmul_config = copy.deepcopy(op_capability['MatMul']) other_config = copy.deepcopy(op_capability['default']) self.quantizable_op_details = OrderedDict() self.recipes_ops = {} self._init_op_stat = {i: [] for i in tf_quantizable_op_type} exclude_first_quantizable_op = (True if (('first_conv_or_matmul_quantization' in self.recipes) and (not self.recipes['first_conv_or_matmul_quantization'])) else False) for details in matched_nodes: node_op = details[(- 1)][0] node_name = details[0] patterns = details[(- 1)] pat_length = len(patterns) pattern_info = {'sequence': [[','.join(patterns[:(pat_length - i)]) for i in range(pat_length)][0]], 'precision': ['int8']} first_conv_or_matmul_node = [] if ((node_op in tf_quantizable_op_type) and (node_name not in self.exclude_node_names) and ((node_name, self.unify_op_type_mapping[node_op]) not in self.quantizable_op_details)): if (((self.unify_op_type_mapping[node_op].find('conv2d') != (- 1)) or (self.unify_op_type_mapping[node_op].find('matmul') != (- 1))) and (len(first_conv_or_matmul_node) == 0)): first_conv_or_matmul_node.append((node_name, self.unify_op_type_mapping[node_op])) self.recipes_ops['first_conv_or_matmul_quantization'] = first_conv_or_matmul_node if (exclude_first_quantizable_op and ((self.unify_op_type_mapping[node_op].find('conv2d') != (- 1)) or (self.unify_op_type_mapping[node_op].find('matmul') != (- 1)))): exclude_first_quantizable_op = False self.exclude_node_names.append(node_name) continue self._init_op_stat[node_op].append(node_name) if (self.unify_op_type_mapping[node_op].find('conv2d') != (- 1)): conv2d_int8_config = copy.deepcopy(conv_config) conv2d_int8_config['pattern'] = pattern_info self.quantizable_op_details[(node_name, self.unify_op_type_mapping[node_op])] = [conv2d_int8_config, fp32_common_config] elif (self.unify_op_type_mapping[node_op].find('conv3d') != (- 1)): conv3d_int8_config = copy.deepcopy(conv3d_config) conv3d_int8_config['pattern'] = pattern_info self.quantizable_op_details[(node_name, self.unify_op_type_mapping[node_op])] = [conv3d_int8_config, fp32_common_config] elif (self.unify_op_type_mapping[node_op].find('matmul') != (- 1)): matmul_int8_config = copy.deepcopy(matmul_config) matmul_int8_config['pattern'] = pattern_info matmul_scheme = ['asym'] matmul_int8_config['activation']['scheme'] = matmul_scheme self.quantizable_op_details[(node_name, self.unify_op_type_mapping[node_op])] = [matmul_int8_config, fp32_common_config] else: self.quantizable_op_details[(node_name, self.unify_op_type_mapping[node_op])] = [copy.deepcopy(other_config), fp32_common_config] if ((node_op in bf16_type) and ((('bf16' in valid_precision) and CpuInfo().bf16) or (os.getenv('FORCE_BF16') == '1'))): self.quantizable_op_details[(node_name, self.unify_op_type_mapping[node_op])].insert(1, bf16_common_config) self.quantize_config['op_wise_config'][node_name] = (False, 'minmax', False) return self.quantizable_op_details def _filter_unquantizable_concat(self, matched_nodes): target_concat_nodes = [i[0] for i in matched_nodes if (i[(- 1)][0] == 'ConcatV2')] from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper from neural_compressor.adaptor.tf_utils.util import GraphAnalyzer g = GraphAnalyzer() g.graph = self.pre_optimized_model.graph_def graph_info = g.parse_graph() concat_nodes = g.query_fusion_pattern_nodes([['ConcatV2']]) for i in concat_nodes: concat_node_name = i[0] if (concat_node_name not in target_concat_nodes): continue input_positive_status = [] for index in range(graph_info[concat_node_name].node.attr['N'].i): each_input_name = GraphRewriterHelper.node_name_from_input(graph_info[concat_node_name].node.input[index]) each_input_node = graph_info[each_input_name].node positive_input = False if (each_input_node.op in ('Relu', 'Relu6')): positive_input = True else: positive_input = g.has_positive_input(each_input_node.name) input_positive_status.append(positive_input) if (not any(input_positive_status)): matched_nodes.remove(i) def _filter_unquantizable_concat_performance_only(self, matched_nodes): target_concat_nodes = [i[0] for i in matched_nodes if (i[(- 1)][0] == 'ConcatV2')] from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper from neural_compressor.adaptor.tf_utils.util import GraphAnalyzer g = GraphAnalyzer() g.graph = self.pre_optimized_model.graph_def graph_info = g.parse_graph() concat_nodes = g.query_fusion_pattern_nodes([['ConcatV2']]) for i in concat_nodes: concat_node_name = i[0] if (concat_node_name not in target_concat_nodes): continue input_positive_status = [] control_flow = False for index in range(graph_info[concat_node_name].node.attr['N'].i): each_input_name = GraphRewriterHelper.node_name_from_input(graph_info[concat_node_name].node.input[index]) each_input_node = graph_info[each_input_name].node if (each_input_node.op in 'Switch'): control_flow = True if control_flow: matched_nodes.remove(i) def query_fw_capability(self, model): if (self.pre_optimized_model is None): from .tf_utils.graph_rewriter.generic.pre_optimize import PreOptimization self.pre_optimizer_handle = PreOptimization(model, self.new_api, self.device) self.pre_optimized_model = self.pre_optimizer_handle.get_optimized_model(self.itex_mode) model.graph_def = self.pre_optimized_model.graph_def self.exclude_node_names = self.pre_optimizer_handle.get_excluded_node_names() patterns = self.query_handler.generate_internal_patterns() bf16_patterns = self.query_handler.get_bf16_patterns() matched_nodes = self.pre_optimizer_handle.get_matched_nodes(patterns) matched_bf16_nodes = self.pre_optimizer_handle.get_matched_nodes(bf16_patterns) original_graph_node_name = [i.name for i in model.graph_def.node] matched_nodes = sorted(matched_nodes, reverse=True, key=(lambda i: (original_graph_node_name.index(i[0]), len(i[(- 1)])))) def check_match(patterns, input_pattern): for i in patterns: if (input_pattern == [i for i in i.replace('+', ' ').strip().split(' ') if i]): return True return False if ((self.new_api and self.performance_only) or self.itex_mode or (os.getenv('TF_FORCE_CONCAT_OPTS') == '1')): self._filter_unquantizable_concat_performance_only(matched_nodes) else: self._filter_unquantizable_concat(matched_nodes) copied_matched_nodes = copy.deepcopy(matched_nodes) for i in copied_matched_nodes: if (i[(- 1)][0] in self.query_handler.get_op_types()['int8']): continue if ((not self.pre_optimizer_handle.has_positive_input(i[0])) and (not check_match(self.query_handler.get_fuse_patterns()['int8'], i[(- 1)]))): matched_nodes.remove(i) del copied_matched_nodes copied_matched_nodes = copy.deepcopy(matched_bf16_nodes) for i in copied_matched_nodes: for j in matched_nodes: if ((i[0] == j[0]) and (i in matched_bf16_nodes)): matched_bf16_nodes.remove(i) del copied_matched_nodes self._query_quantizable_ops(matched_nodes) self._query_bf16_ops(matched_bf16_nodes) capability = {'optypewise': self.get_optype_wise_ability(), 'recipes_ops': self.recipes_ops} capability['opwise'] = copy.deepcopy(self.quantizable_op_details) capability['opwise'].update(self.bf16_op_details) logger.debug('Dump framework quantization capability:') logger.debug(capability) return capability def set_tensor(self, model, tensor_dict): from .tf_utils.graph_util import GraphAnalyzer g = GraphAnalyzer() g.graph = model.graph_def graph_info = g.parse_graph() def _get_fp32_op_name(model, tensor_name): is_weight = False is_biasadd = False last_node_name = None current_node_name = None for each_node in model.graph_def.node: if (tensor_name in each_node.input): tensor_index = list(each_node.input).index(tensor_name) if ((each_node.op.find('Quantized') != (- 1)) and (tensor_index == 2)): is_biasadd = True last_node_name = each_node.input[0] current_node_name = each_node.name if ((tensor_name + '_qint8_const') in each_node.input): pass return (is_weight, is_biasadd, current_node_name, last_node_name) from tensorflow.core.framework import attr_value_pb2 from tensorflow.python.framework import dtypes, tensor_util from neural_compressor.adaptor.tf_utils.graph_util import GraphRewriterHelper as Helper qint32_type = dtypes.qint32.as_datatype_enum for (tensor_name, tensor_content) in tensor_dict.items(): (is_weight, is_biasadd, current_node_name, last_node_name) = _get_fp32_op_name(model, tensor_name) if is_biasadd: is_biasadd_dtype_is_fp32 = (graph_info[current_node_name].node.attr['Tbias'] == attr_value_pb2.AttrValue(type=dtypes.float32.as_datatype_enum)) current_node = graph_info[current_node_name].node bias_add_node = graph_info[current_node.input[2]].node if is_biasadd_dtype_is_fp32: bias_add_node.attr['value'].CopyFrom(attr_value_pb2.AttrValue(tensor=tensor_util.make_tensor_proto(tensor_content, dtypes.float32, tensor_content.shape))) else: last_node = graph_info[last_node_name].node min_input = graph_info[last_node.input[(- 2)]].node.attr['value'].tensor.float_val[0] max_input = graph_info[last_node.input[(- 1)]].node.attr['value'].tensor.float_val[0] channel_size = tensor_content.shape[0] max_filter_node = graph_info[current_node.input[6]].node min_filter_node = graph_info[current_node.input[5]].node if max_filter_node.attr['value'].tensor.float_val: max_filter_tensor = [] min_filter_tensor = [] max_filter_tensor.append(max_filter_node.attr['value'].tensor.float_val[0]) min_filter_tensor.append(min_filter_node.attr['value'].tensor.float_val[0]) else: max_filter_tensor = tensor_util.MakeNdarray(min_filter_node.attr['value'].tensor) min_filter_tensor = tensor_util.MakeNdarray(min_filter_node.attr['value'].tensor) activation_range = (127.0 if (current_node.attr['Tinput'].type == dtypes.qint8) else 255.0) updated_bias = Helper.generate_int32_bias_for_conv(tensor_content, channel_size, max_input, min_input, max_filter_tensor, min_filter_tensor, activation_range) bias_add_node.attr['dtype'].CopyFrom(attr_value_pb2.AttrValue(type=qint32_type)) bias_add_node.attr['value'].CopyFrom(attr_value_pb2.AttrValue(tensor=tensor_util.make_tensor_proto(updated_bias, dtypes.int32, tensor_content.shape))) bias_add_node.attr['value'].tensor.dtype = qint32_type current_node.attr['Tbias'].CopyFrom(attr_value_pb2.AttrValue(type=qint32_type)) if is_weight: tmp_const_node = Helper.create_constant_node((current_node.name + '_weights_tmp'), tensor_content.transpose(2, 3, 1, 0), dtypes.float32) min_filter_node = graph_info[current_node.input[5]].node per_channel = (True if min_filter_node.attr['value'].tensor.tensor_shape else False) from .tf_utils.quantize_graph_common import QuantizeGraphHelper original_fp32_op = current_node.op.split('With')[0].split('Quantized')[(- 1)] if (original_fp32_op.find('Depthwise') != (- 1)): original_fp32_op = 'DepthwiseConv2dNative' (qint8_const_node, min_node, max_node) = QuantizeGraphHelper.generate_quantized_weight_node(original_fp32_op, tmp_const_node, per_channel) g.add_node(qint8_const_node, [], [current_node.name]) g.add_node(min_node, [], [current_node.name]) g.add_node(max_node, [], [current_node.name]) g.replace_constant_graph_with_constant_node(qint8_const_node, tensor_name) g.replace_constant_graph_with_constant_node(min_node, current_node.input[5]) g.replace_constant_graph_with_constant_node(max_node, current_node.input[6]) def inspect_weight_and_bias(self, node_list, graph_def, graph_info, graph_node_name_mapping): import tensorflow as tf from neural_compressor.adaptor.tf_utils.util import get_tensor_val_from_graph_node from neural_compressor.utils.utility import dequantize_weight from .tf_utils.util import int8_node_name_reverse weights_result = {} inspect_nodes = [] node_set = set(node_list) for node in graph_def.node: node_name = node.name if ('Quantized' in node.op): node_name = int8_node_name_reverse(node) if ((node_name in node_set) and (('Conv' in node.op) or ('Mul' in node.op))): inspect_nodes.append(node) logger.debug(f'Start to inspect weight and bias for: {[node.name for node in inspect_nodes]}.') for node in inspect_nodes: node_name = node.name weight_node_name = node.input[1] weight_node = graph_node_name_mapping[weight_node_name] if (weight_node.op != 'Const'): continue weight_node_val = get_tensor_val_from_graph_node(graph_node_name_mapping, weight_node_name) weight_node_val = weight_node_val.astype('float32') if ('Quantized' in node.op): node_name = int8_node_name_reverse(node) weight_node_name_pre = weight_node_name.split('_qint8_const')[0] min_filter_node = (weight_node_name_pre + '_min') max_filter_node = (weight_node_name_pre + '_max') if graph_info[min_filter_node].node.attr['value'].tensor.float_val: min_filter_val = graph_info[min_filter_node].node.attr['value'].tensor.float_val max_filter_val = graph_info[max_filter_node].node.attr['value'].tensor.float_val else: min_filter_val = get_tensor_val_from_graph_node(graph_node_name_mapping, min_filter_node) max_filter_val = get_tensor_val_from_graph_node(graph_node_name_mapping, max_filter_node) weight_node_val = dequantize_weight(weight_node_val, min_filter_val, max_filter_val) weights_result[node_name] = {weight_node_name: weight_node_val} return weights_result def fused_node_mapping(self, node_list, pattern_mapping, graph_info, graph_node_name_mapping): fused_mapping = {} fused_mapping_reverse = {} for node_name in node_list: fused_seq = pattern_mapping[node_name]['sequence'].split(',') if (len(fused_seq) == 1): fused_mapping[node_name] = node_name fused_mapping_reverse[node_name] = node_name continue _next_node_name = node_name for _next_node_op_type in fused_seq[1:]: node_details = graph_info[_next_node_name] for node_output_name in node_details.outputs: if (graph_node_name_mapping[node_output_name].op == 'Cast'): cast_node = graph_node_name_mapping[node_output_name] node_output_name = graph_info[cast_node.name].outputs[0] if (graph_node_name_mapping[node_output_name].op in [_next_node_op_type, 'Cast']): _next_node_name = node_output_name fused_mapping[node_name] = _next_node_name fused_mapping_reverse[_next_node_name] = node_name return (fused_mapping, fused_mapping_reverse) def _inspect_tensor_inference(self, inspect_node_dict, model, dataloader, iteration_list): out_tensor_lst = [] out_tensor_lst += [{n: [((n + ':') + str(i)) for i in range(3)]} for n in inspect_node_dict['qreq_node']] out_tensor_lst += [{n: (n + ':0')} for n in inspect_node_dict['qdq_node']] out_tensor_lst += [{n: (n + ':0')} for n in inspect_node_dict['f_node']] out_cnt = len(out_tensor_lst) iteration_list = set(iteration_list) input_tensor = model.input_tensor logger.info('Start to do inference for inspect activation.') activation_result = [] for (idx, (inputs, labels)) in enumerate(dataloader): model_out = [] if ((idx + 1) > max(iteration_list)): break if ((idx + 1) not in iteration_list): continue if (len(input_tensor) == 1): feed_dict = {input_tensor[0]: inputs} else: assert (len(input_tensor) == len(inputs)), 'inputs len must equal with input_tensor' feed_dict = dict(zip(input_tensor, inputs)) for (i, out_t) in enumerate(out_tensor_lst): logger.debug(f'Finished inspect {i}/{out_cnt} nodes, current inspect node {out_t.keys()}.') model_out.append(model.sess.run(out_t, feed_dict)) activation_result.append(model_out) return activation_result def inspect_activation(self, node_list, graph_def, graph_node_name_mapping, quantization_cfg, dataloader, iteration_list, graph_info): from neural_compressor.model import Model original_graph_node_mapping = {} for node in graph_def.node: original_graph_node_mapping[node.name] = node inspect_node_dict = {'qdq_node': [], 'qreq_node': [], 'f_node': []} for node_name in node_list: node = graph_node_name_mapping[node_name] if (('Quantized' in node.op) and ('Dequantize' in node.op)): inspect_node_dict['qdq_node'].append(node.name) elif (('Quantized' in node.op) or ('_Quantized' in node.op) or ('Requantize' in node.op)): inspect_node_dict['qreq_node'].append(node.name) else: inspect_node_dict['f_node'].append(node_name) pattern_mapping = {} node_dict = quantization_cfg['op'] for node_name_and_type in node_dict.keys(): (node_name, _) = node_name_and_type if ('pattern' in node_dict[node_name_and_type]): pattern_mapping[node_name] = node_dict[node_name_and_type]['pattern'] else: pattern_mapping[node_name] = {'sequence': node_name} if inspect_node_dict['f_node']: (fuse_map, fuse_map_reverse) = self.fused_node_mapping(inspect_node_dict['f_node'], pattern_mapping, graph_info, graph_node_name_mapping) inspect_node_dict['f_node'] = [fuse_map[n] for n in inspect_node_dict['f_node']] model = Model(graph_def) activation_result = self._inspect_tensor_inference(inspect_node_dict, model, dataloader, iteration_list) final_result = [] int8_postfix = '_eightbit' for iter_res in activation_result: tmp_iter_result = {} for res in iter_res: (node_name, val) = (list(res.keys())[0], list(res.values())[0]) val = (Dequantize(val[0], (node_name, val[1], val[2])) if (len(val) == 3) else val) val = val.astype(np.float32) index_postfix = node_name.find(int8_postfix) if (index_postfix != (- 1)): node_name = node_name[:index_postfix] tmp_iter_result[node_name] = {node_name: val} else: tmp_iter_result[fuse_map_reverse[node_name]] = {fuse_map_reverse[node_name]: val} final_result.append(tmp_iter_result) return final_result def inspect_tensor(self, model, dataloader=None, op_list=[], iteration_list=[], inspect_type='activation', save_to_disk=False, save_path=None, quantization_cfg=None): import tensorflow as tf from neural_compressor.adaptor.tf_utils.graph_util import GraphAnalyzer from neural_compressor.model.tensorflow_model import TensorflowBaseModel from neural_compressor.utils.utility import dump_data_to_local, load_data_from_pkl from .tf_utils.util import int8_node_name_reverse if isinstance(model, TensorflowBaseModel): model = model.graph_def if (not quantization_cfg): quantization_cfg = load_data_from_pkl('./nc_workspace/', 'cfg.pkl') node_list = op_list graph_node_name_mapping = {} quan_model_flag = False for node in model.node: node_name = int8_node_name_reverse(node) if ('Quantized' in node.op): quan_model_flag = True node_name = int8_node_name_reverse(node) if (node.attr['value'].tensor.dtype == tf.dtypes.bfloat16.as_datatype_enum): quan_model_flag = True graph_node_name_mapping[node_name] = node if quan_model_flag: logger.info('Dump the tensor for quantized model.') g = GraphAnalyzer() g.graph = model graph_info = g.parse_graph() inspect_result = {} if ((inspect_type == 'weight') or (inspect_type == 'all')): logger.info('Start to inspect weight and bias.') weights_result = self.inspect_weight_and_bias(node_list, model, graph_info, graph_node_name_mapping) inspect_result['weight'] = weights_result if ((inspect_type == 'activation') or (inspect_type == 'all')): logger.info('Start to inspect activation.') activation_result = self.inspect_activation(node_list, model, graph_node_name_mapping, quantization_cfg, dataloader, iteration_list, graph_info) inspect_result['activation'] = activation_result if save_to_disk: if (not save_path): save_path = './nc_workspace/tmp/' dump_data_to_local(inspect_result, save_path, 'inspect_result.pkl') logger.info(f'Dumped the inspect tensor to {save_path}') return inspect_result def quantize_input(self, model): scale = None import tensorflow as tf if version1_lt_version2(tf.version.VERSION, '2.1.0'): logger.warning('Quantize input needs tensorflow 2.1.0 and newer.') return (model, scale) graph_def = model.as_graph_def() node_name_mapping = {} quantize_nodes = [] for node in graph_def.node: node_name_mapping[node.name] = node if (node.op == 'QuantizeV2'): quantize_nodes.append(node) target_quantize_nodes = [] for node in quantize_nodes: if (((node_name_mapping[node.input[0]].op == 'Pad') and (node_name_mapping[node_name_mapping[node.input[0]].input[0]].op == 'Placeholder')) or (node_name_mapping[node.input[0]].op == 'Placeholder')): target_quantize_nodes.append(node) assert (len(target_quantize_nodes) == 1), 'only support 1 QuantizeV2 from Placeholder' quantize_node = target_quantize_nodes[0] quantize_node_input = node_name_mapping[quantize_node.input[0]] quantize_node_outputs = [node for node in graph_def.node if (quantize_node.name in node.input)] from .tf_utils.graph_util import GraphRewriterHelper if (quantize_node_input.op == 'Pad'): pad_node_input = node_name_mapping[quantize_node_input.input[0]] assert (pad_node_input.op == 'Placeholder'), 'only support Pad between QuantizeV2 and Placeholder' from tensorflow.python.framework import tensor_util paddings_tensor = tensor_util.MakeNdarray(node_name_mapping[quantize_node_input.input[1]].attr['value'].tensor).flatten() quantize_node.input[0] = quantize_node_input.input[0] for conv_node in quantize_node_outputs: assert ('Conv2D' in conv_node.op), 'only support QuantizeV2 to Conv2D' GraphRewriterHelper.set_attr_int_list(conv_node, 'padding_list', paddings_tensor) graph_def.node.remove(quantize_node_input) from tensorflow.python.framework import dtypes GraphRewriterHelper.set_attr_dtype(node_name_mapping[quantize_node.input[0]], 'dtype', dtypes.qint8) for conv_node in quantize_node_outputs: for (index, conv_input) in enumerate(conv_node.input): if (conv_input == quantize_node.name): conv_node.input[index] = quantize_node.input[0] elif (conv_input == (quantize_node.name + ':1')): conv_node.input[index] = quantize_node.input[1] elif (conv_input == (quantize_node.name + ':2')): conv_node.input[index] = quantize_node.input[2] max_node = node_name_mapping[quantize_node.input[2]] min_node = node_name_mapping[quantize_node.input[1]] max_value = max_node.attr['value'].tensor.float_val[0] min_value = min_node.attr['value'].tensor.float_val[0] scale = (127.0 / max(abs(max_value), abs(min_value))) graph_def.node.remove(quantize_node) graph = tensorflow.Graph() with graph.as_default(): tensorflow.import_graph_def(graph_def, name='') return (graph, scale) def get_optype_wise_ability(self): res = OrderedDict() for op in self.quantizable_op_details: if (op[1] not in res): res[op[1]] = {'activation': self.quantizable_op_details[op][0]['activation']} if ('weight' in self.quantizable_op_details[op][0]): res[op[1]]['weight'] = self.quantizable_op_details[op][0]['weight'] for op in self.bf16_op_details: if (op[1] not in res): res[op[1]] = {'activation': {'dtype': ['bf16']}, 'weight': {'dtype': ['bf16']}} return res def _pre_hook_for_qat(self, dataloader=None): self.model.model = self.qat_convert(self.model.model) def _post_hook_for_qat(self): pass def _pre_eval_hook(self, model): return model def _post_eval_hook(self, model, **kwargs): pass def save(self, model, path): pass def convert(self, model, source, destination): assert ((source.lower() == 'qat') and (destination.lower() == 'default')) capability = self.query_fw_capability(model) quantize_config = {'op_wise_config': {}} for each_op_info in capability['opwise']: is_perchannel = False weight_bit = 7.0 for op_cap in capability['opwise'][each_op_info]: if (('activation' in op_cap) and ('quant_mode' in op_cap['activation'])): activation = op_cap['activation'] if ('weight' in op_cap): weight = op_cap['weight'] is_perchannel = (True if (weight['granularity'][0] == 'per_channel') else False) algorithm = activation['algorithm'][0] is_asymmetric = False if ('activation' in op_cap): is_asymmetric = (True if (activation['scheme'][0] == 'asym') else False) quantize_config['op_wise_config'][each_op_info[0]] = (is_perchannel, algorithm, is_asymmetric, weight_bit) from .tf_utils.graph_converter import GraphConverter tmp_graphdef = copy.deepcopy(model.graph_def) for i in tmp_graphdef.node: if ((i.op == 'Const') and i.input): i.ClearField('input') model.graph_def = tmp_graphdef converter = GraphConverter(model, qt_config=quantize_config, int8_sequences=self.op_wise_sequences, fake_quant=True, new_api=self.new_api, performance_only=self.performance_only, use_bf16=self.use_bf16) return converter.convert() def qat_convert(self, model, quantize_recipe=None): import tensorflow as tf assert isinstance(model, tf.keras.Model), 'The model to be converted is expected to be a `tf.keras.Model` instance. You should not pass an instance of type: {input}.'.format(input=model.__class__.__name__) assert (model.__class__.__name__ in ['Functional', 'Sequential']), 'Only `Functional` or `Sequential` keras model is supported for QAT.' from .tf_utils.quantize_graph.qat.quantize_config import global_config from .tf_utils.quantize_graph.qat.quantize_helper import init_quantize_config, qat_clone_function config = init_quantize_config(model, quantize_recipe) q_model = tf.keras.models.clone_model(model, input_tensors=None, clone_function=qat_clone_function) global_config.clear() return q_model _elapsed_time('Pass recover model') def recover_tuned_model(self, model, q_config): from .tf_utils.graph_rewriter.generic.pre_optimize import PreOptimization self.pre_optimizer_handle = PreOptimization(model, self.new_api, self.device) self.pre_optimized_model = self.pre_optimizer_handle.get_optimized_model(self.itex_mode) model.graph_def = self.pre_optimized_model.graph_def from .tf_utils.graph_converter_without_calib import GraphConverterWithoutCalib converter = GraphConverterWithoutCalib(model, recover_config=q_config, new_api=self.new_api, performance_only=self.performance_only, use_bf16=self.use_bf16) return converter.convert_without_calib() def diagnosis_helper(self, fp32_model, quan_model, tune_cfg, save_path): from .tf_utils.util import tf_diagnosis_helper return tf_diagnosis_helper(fp32_model, quan_model, tune_cfg, save_path) def get_output_op_names(self, qmodel): from .tf_utils.graph_util import GraphAnalyzer graph_def = GraphAnalyzer().parse_graph(qmodel.graph_def) output_op_names = set() def _add_output_op_name(opname): if opname.endswith('_dequantize'): output_op_names.add(opname[:(- len('_dequantize'))]) elif opname.endswith('__dequant'): pass else: output_op_names.add(opname) for output_opname in qmodel.output_node_names: op_count = 0 stack = [output_opname] while stack: opname = stack.pop() while True: op_count += 1 if (opname not in graph_def): break op = graph_def[opname] if (op.node.op == 'Dequantize'): _add_output_op_name(opname) break next_opnames = op.node.input if (not next_opnames): break elif (len(next_opnames) > 1): stack += next_opnames[1:] opname = next_opnames[0] output_op_names = list(output_op_names) logger.debug(f'output op names: {output_op_names}') return output_op_names def calculate_op_sensitivity(self, model, dataloader, tune_cfg, output_op_names, confidence_batches, fallback=True, requantize_cfgs=None): from copy import deepcopy fp32_op_cfg = {'activation': {'dtype': 'fp32', 'quant_mode': 'fp32'}, 'weight': {'dtype': 'fp32'}} if fallback: ops_list = [op for (op, config) in tune_cfg['op'].items() if (config['activation']['quant_mode'] in ('static', 'dynamic'))] replace_cfgs = {op: fp32_op_cfg for op in tune_cfg['op']} else: ops_list = [op for (op, config) in tune_cfg['op'].items() if ((config['activation']['quant_mode'] == 'fp32') and (op in requantize_cfgs))] replace_cfgs = requantize_cfgs mse_result = self._get_mse_order(model, deepcopy(tune_cfg), replace_cfgs, ops_list, dataloader, output_op_names, confidence_batches) mse_order = [op for (op, _) in sorted(mse_result.items(), key=(lambda i: i[1]))] logger.debug('Dump MSE order:') for op in mse_order: logger.debug(f'{op}: {mse_result[op]}') return mse_order def _get_mse_order(self, fp32_model, tune_cfg, replace_cfgs, ops_lst, dataloader, output_op_names, confidence_batches): op_cfg = tune_cfg['op'] mse_result = {} partial_dataloader = self._partial_dataloader(dataloader, confidence_batches) fp32_output = self._inference_model_on_batches(fp32_model, tune_cfg, partial_dataloader, output_op_names) for op in ops_lst: backup_cfg = op_cfg[op] op_cfg[op] = replace_cfgs[op] q_model = self.quantize(tune_cfg, fp32_model, partial_dataloader) q_output = self._inference_model_on_batches(q_model, tune_cfg, partial_dataloader, output_op_names) mse_result[op] = self._calculate_mse(fp32_output, q_output) op_cfg[op] = backup_cfg return mse_result def _partial_dataset_of(self, dataloader, confidence_batches): from neural_compressor.data.datasets.dummy_dataset import DummyDataset from neural_compressor.data.datasets.dummy_dataset import DummyDataset as DummyDataset_v2_x if (isinstance(dataloader.dataset, DummyDataset) or isinstance(dataloader.dataset, DummyDataset_v2_x)): assert isinstance(confidence_batches, int) ds = copy.deepcopy(dataloader.dataset) ds.dataset = ds.dataset[:confidence_batches] return ds else: return dataloader.dataset.take(confidence_batches) def _partial_dataloader(self, dataloader, confidence_batches): return type(dataloader)(dataset=self._partial_dataset_of(dataloader, confidence_batches), batch_size=dataloader.batch_size, last_batch=dataloader.last_batch, collate_fn=dataloader.collate_fn, sampler=dataloader.sampler, batch_sampler=dataloader.batch_sampler, num_workers=dataloader.num_workers, pin_memory=dataloader.pin_memory, shuffle=dataloader.shuffle, distributed=dataloader.distributed) def _calculate_mse(self, fp32_output, q_output): result = [] for (i, j) in zip(fp32_output, q_output): result.append(np.square((i - j)).mean()) return np.array(result).mean() def _inference_model_on_batches(self, model, tune_cfg, dataloader, output_op_names): from .tf_utils.util import generate_feed_dict input_tensors = model.input_tensor output_tensors = [] for op in output_op_names: for tensor in model.graph.get_operation_by_name(op).outputs: output_tensors.append(tensor) predictions = [] for (index, (inputs, _)) in enumerate(dataloader): feed_dict = generate_feed_dict(input_tensors, inputs) pred = model.sess.run(output_tensors, feed_dict) for item in pred: predictions.append(item) return predictions def smooth_quant(self, model, dataloader, calib_iter=1, alpha=0.5, folding=False, percentile=99.999, op_types=['MatMul', 'Conv2D'], scales_per_op=True, record_max_info=False, weight_clip=True, auto_alpha_args={'alpha_min': 0.0, 'alpha_max': 1.0, 'alpha_step': 0.1, 'shared_criterion': 'mean', 'do_blockwise': False}, default_alpha=0.5): logger.info('Start Smoothing process for Smooth Quantization.') if (self.smooth_quant_model is not None): return self.smooth_quant_model if (model.model_type == 'llm_saved_model'): return self.smooth_quant_LLM(model, dataloader, calib_iter, alpha, folding, percentile, op_types, scales_per_op) from .tf_utils.graph_rewriter.generic.pre_optimize import PreOptimization self.pre_optimizer_handle = PreOptimization(model, self.new_api, self.device) self.pre_optimized_model = self.pre_optimizer_handle.get_optimized_model(self.itex_mode) model.graph_def = self.pre_optimized_model.graph_def from .tf_utils.smooth_quant_calibration import SmoothQuantCalibration calibration = SmoothQuantCalibration(model, dataloader, calib_iter, op_types, percentile) (max_vals_per_channel, sq_weight_node_names) = calibration() from neural_compressor.adaptor.tf_utils.util import get_weight_from_input_tensor (sq_weight_tensors, sq_weights_nodes) = get_weight_from_input_tensor(model, max_vals_per_channel.keys(), op_types) from .tf_utils.smooth_quant_scaler import SmoothQuantScaler scaler = SmoothQuantScaler(model, dataloader, alpha, scales_per_op) (model, mul_list) = scaler.transform(max_vals_per_channel, sq_weight_tensors, sq_weights_nodes, sq_weight_node_names) self.smooth_quant_mul_ops.extend(mul_list) self.smooth_quant_model = model return self.smooth_quant_model def smooth_quant_LLM(self, model, dataloader, calib_iter=1, alpha=0.5, folding=False, percentile=99.999, op_types=['MatMul', 'Conv2D'], scales_per_op=True): from .tf_utils.graph_rewriter.generic.pre_optimize import PreOptimization self.pre_optimizer_handle = PreOptimization(model, self.new_api, self.device) self.pre_optimized_model = self.pre_optimizer_handle.get_optimized_model(self.itex_mode) model.graph_def = self.pre_optimized_model.graph_def op_types = ['MatMul'] llm_temp_dir = (self.work_dir + '/temp_saved_model') from .tf_utils.smooth_quant_calibration import SmoothQuantCalibrationLLM calibration = SmoothQuantCalibrationLLM(model._model, dataloader, calib_iter, op_types, percentile, llm_temp_dir, model.weight_name_mapping) (max_vals_per_channel, sq_target_node_names, sq_weight_tensor_dict, sq_graph_def) = calibration(model.input_node_names, model.output_node_names) from .tf_utils.smooth_quant_scaler import SmoothQuantScalerLLM scaler = SmoothQuantScalerLLM(sq_graph_def, alpha, scales_per_op, op_types) (sq_graph_def, sq_weight_scale_dict, mul_list) = scaler.transform(max_vals_per_channel, sq_weight_tensor_dict, sq_target_node_names) model.graph_def = sq_graph_def model.model_path = llm_temp_dir model.sq_weight_scale_dict = sq_weight_scale_dict self.smooth_quant_mul_ops.extend(mul_list) self.smooth_quant_model = model return self.smooth_quant_model
def seed(seed): random.seed(seed) numpy.random.seed(seed) torch.manual_seed(seed) if torch.cuda.is_available(): torch.cuda.manual_seed_all(seed)
def train(args): transform = transforms.Compose([transforms.RandomCrop(args.crop_size), transforms.RandomHorizontalFlip(), transforms.ToTensor(), transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))]) transform_dev = transforms.Compose([transforms.CenterCrop(args.crop_size), transforms.ToTensor(), transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))]) vocab = Vocabulary() vocab.load(DICT.format(args.data_set)) data_loader = get_loader(IMAGE_ROOT.format(args.data_set), CAPT.format(args.data_set, 'train'), vocab, transform, args.batch_size, shuffle=True, return_target=True, num_workers=args.num_workers) data_loader_dev = get_loader(IMAGE_ROOT.format(args.data_set), CAPT.format(args.data_set, 'val'), vocab, transform_dev, args.batch_size, shuffle=False, return_target=True, num_workers=args.num_workers) ranker = Ranker(root=IMAGE_ROOT.format(args.data_set), image_split_file=SPLIT.format(args.data_set, 'val'), transform=transform_dev, num_workers=args.num_workers) save_folder = '{}/{}-{}'.format(args.save, args.data_set, time.strftime('%Y%m%d-%H%M%S')) create_exp_dir(save_folder, scripts_to_save=['models.py', 'data_loader.py', 'train.py', 'build_vocab.py', 'utils.py']) def logging(s, print_=True, log_=True): if print_: print(s) if log_: with open(os.path.join(save_folder, 'log.txt'), 'a+') as f_log: f_log.write((s + '\n')) logging(str(args)) image_encoder = DummyImageEncoder(args.embed_size).to(device) caption_encoder = DummyCaptionEncoder(vocab_size=len(vocab), vocab_embed_size=(args.embed_size * 2), embed_size=args.embed_size).to(device) image_encoder.train() caption_encoder.train() params = (image_encoder.get_trainable_parameters() + caption_encoder.get_trainable_parameters()) current_lr = args.learning_rate optimizer = torch.optim.Adam(params, lr=current_lr) cur_patient = 0 best_score = float('-inf') stop_train = False total_step = len(data_loader) for epoch in range(1): for (i, (target_images, candidate_images, captions, lengths, meta_info)) in enumerate(data_loader): target_images = target_images.to(device) target_ft = image_encoder.forward(target_images) candidate_images = candidate_images.to(device) candidate_ft = image_encoder.forward(candidate_images) captions = captions.to(device) caption_ft = caption_encoder(captions, lengths) m = target_images.size(0) random_index = [((m - 1) - n) for n in range(m)] random_index = torch.LongTensor(random_index) negative_ft = target_ft[random_index] loss = triplet_avg(anchor=(candidate_ft + caption_ft), positive=target_ft, negative=negative_ft) caption_encoder.zero_grad() image_encoder.zero_grad() loss.backward() optimizer.step() if ((i % args.log_step) == 0): logging('| epoch {:3d} | step {:6d}/{:6d} | lr {:06.6f} | train loss {:8.3f}'.format(epoch, i, total_step, current_lr, loss.item())) image_encoder.eval() caption_encoder.eval() logging(('-' * 77)) metrics = eval_batch(data_loader_dev, image_encoder, caption_encoder, ranker) logging('| eval loss: {:8.3f} | score {:8.5f} / {:8.5f} '.format(metrics['loss'], metrics['score'], best_score)) logging(('-' * 77)) image_encoder.train() caption_encoder.train() dev_score = metrics['score'] if (dev_score > best_score): best_score = dev_score resnet = image_encoder.delete_resnet() torch.save(image_encoder.state_dict(), os.path.join(save_folder, 'image-{}.th'.format(args.embed_size))) image_encoder.load_resnet(resnet) torch.save(caption_encoder.state_dict(), os.path.join(save_folder, 'cap-{}.th'.format(args.embed_size))) cur_patient = 0 else: cur_patient += 1 if (cur_patient >= args.patient): current_lr /= 2 for param_group in optimizer.param_groups: param_group['lr'] = current_lr if (current_lr < (args.learning_rate * 0.001)): stop_train = True break if stop_train: break logging('best_dev_score: {}'.format(best_score))
def add_eval_lm_args(parser): group = parser.add_argument_group('LM Evaluation') add_common_eval_args(group) group.add_argument('--output-word-probs', action='store_true', help='if set, outputs words and their predicted log probabilities to standard output') group.add_argument('--output-word-stats', action='store_true', help='if set, outputs word statistics such as word count, average probability, etc') group.add_argument('--target-idx', type=int, default=0, help='if set and the language model has multiple targets, evaluates language model for this particular target')
class TestCleaner(): def cleaner(self) -> Cleaner: generator = DummyGenerator() return Cleaner(generator=generator) def key(self) -> chex.PRNGKey: return jax.random.PRNGKey(0) def test_cleaner__reset_jit(self, cleaner: Cleaner) -> None: chex.clear_trace_counter() reset_fn = jax.jit(chex.assert_max_traces(cleaner.reset, n=1)) key = jax.random.PRNGKey(0) (state, timestep) = reset_fn(key) (state, timestep) = reset_fn(key) assert isinstance(timestep, TimeStep) assert isinstance(state, State) def test_cleaner__reset(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: reset_fn = jax.jit(cleaner.reset) (state, timestep) = reset_fn(key) assert isinstance(timestep, TimeStep) assert isinstance(state, State) assert jnp.all((state.agents_locations == jnp.zeros((cleaner.num_agents, 2)))) assert (jnp.sum((state.grid == CLEAN)) == 1) assert (state.step_count == 0) assert_is_jax_array_tree(state) def test_cleaner__step_jit(self, cleaner: Cleaner) -> None: key = jax.random.PRNGKey(0) (state, timestep) = cleaner.reset(key) action = jnp.array([1, 2, 3], jnp.int32) chex.clear_trace_counter() step_fn = jax.jit(chex.assert_max_traces(cleaner.step, n=1)) (next_state, next_timestep) = step_fn(state, action) (next_state, next_timestep) = step_fn(state, action) assert isinstance(next_timestep, TimeStep) assert isinstance(next_state, State) def test_cleaner__step(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: (initial_state, timestep) = cleaner.reset(key) step_fn = jax.jit(cleaner.step) actions = jnp.array(([1] * cleaner.num_agents)) (state, timestep) = step_fn(initial_state, actions) assert (jnp.sum((state.grid != initial_state.grid)) == 1) assert (state.grid[(0, 1)] == CLEAN) assert (timestep.reward == (1 - cleaner.penalty_per_timestep)) assert jnp.all((state.agents_locations == jnp.array([0, 1]))) actions = jnp.array([2, 3, 2]) (state, timestep) = step_fn(state, actions) assert (jnp.sum((state.grid != initial_state.grid)) == 2) assert (state.grid[(0, 1)] == CLEAN) assert (state.grid[(1, 1)] == CLEAN) assert (timestep.reward == (1 - cleaner.penalty_per_timestep)) assert (timestep.step_type == StepType.MID) assert jnp.all((state.agents_locations[0] == jnp.array([1, 1]))) assert jnp.all((state.agents_locations[1] == jnp.array([0, 0]))) assert jnp.all((state.agents_locations[2] == jnp.array([1, 1]))) def test_cleaner__step_invalid_action(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: (state, _) = cleaner.reset(key) step_fn = jax.jit(cleaner.step) actions = jnp.array([0, 1, 1]) (state, timestep) = step_fn(state, actions) assert (timestep.step_type == StepType.LAST) assert jnp.all((state.agents_locations[0] == jnp.array([0, 0]))) assert jnp.all((state.agents_locations[1] == jnp.array([0, 1]))) assert jnp.all((state.agents_locations[2] == jnp.array([0, 1]))) assert (timestep.reward == (1 - cleaner.penalty_per_timestep)) def test_cleaner__initial_action_mask(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: (state, _) = cleaner.reset(key) expected_action_mask = jnp.array([[False, True, False, False] for _ in range(cleaner.num_agents)]) assert jnp.all((state.action_mask == expected_action_mask)) action_mask = cleaner._compute_action_mask(state.grid, state.agents_locations) assert jnp.all((action_mask == expected_action_mask)) def test_cleaner__action_mask(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: (state, _) = cleaner.reset(key) agents_locations = jnp.array([[1, 1], [2, 2], [4, 4]]) action_mask = cleaner._compute_action_mask(state.grid, agents_locations) assert jnp.all((action_mask[0] == jnp.array([True, False, True, False]))) assert jnp.all((action_mask[1] == jnp.array([False, True, False, True]))) assert jnp.all((action_mask[2] == jnp.array([False, False, False, True]))) def test_cleaner__does_not_smoke(self, cleaner: Cleaner) -> None: def select_actions(key: chex.PRNGKey, observation: Observation) -> chex.Array: def select_action(key: chex.PRNGKey, agent_action_mask: chex.Array) -> chex.Array: return jax.random.choice(key, jnp.arange(4), p=agent_action_mask.flatten()) subkeys = jax.random.split(key, cleaner.num_agents) return select_action(subkeys, observation.action_mask) check_env_does_not_smoke(cleaner, select_actions) def test_cleaner__compute_extras(self, cleaner: Cleaner, key: chex.PRNGKey) -> None: (state, _) = cleaner.reset(key) extras = cleaner._compute_extras(state) assert (list(extras.keys()) == ['ratio_dirty_tiles', 'num_dirty_tiles']) assert (0 <= extras['ratio_dirty_tiles'] <= 1) grid = state.grid assert (extras['ratio_dirty_tiles'] == (jnp.sum((grid == DIRTY)) / jnp.sum((grid != WALL)))) assert (extras['num_dirty_tiles'] == jnp.sum((grid == DIRTY)))
class SigmoidFocalClassificationLoss(nn.Module): def __init__(self, gamma: float=2.0, alpha: float=0.25): super(SigmoidFocalClassificationLoss, self).__init__() self.alpha = alpha self.gamma = gamma def sigmoid_cross_entropy_with_logits(input: torch.Tensor, target: torch.Tensor): loss = ((torch.clamp(input, min=0) - (input * target)) + torch.log1p(torch.exp((- torch.abs(input))))) return loss def forward(self, input: torch.Tensor, target: torch.Tensor, weights: torch.Tensor): pred_sigmoid = torch.sigmoid(input) alpha_weight = ((target * self.alpha) + ((1 - target) * (1 - self.alpha))) pt = ((target * (1.0 - pred_sigmoid)) + ((1.0 - target) * pred_sigmoid)) focal_weight = (alpha_weight * torch.pow(pt, self.gamma)) bce_loss = self.sigmoid_cross_entropy_with_logits(input, target) loss = (focal_weight * bce_loss) weights = weights.unsqueeze((- 1)) assert (weights.shape.__len__() == loss.shape.__len__()) return (loss * weights)
def check_collisions(): defining_modules = defaultdict(list) for module in _MODEL_TAG_MODULES: name = module.__name__ for tag in extract_tags(module): defining_modules[tag].append(name) duplicate_defs = {k: v for (k, v) in defining_modules.items() if (len(v) > 1)} for key in duplicate_defs.keys(): print('Variable {} defined multiple times'.format(key))
def flatten(xs: Iterable[Any]) -> List[Any]: return sum((([x] if (not isinstance(x, Iterable)) else flatten(x)) for x in xs), [])
def check_existing_pt_files(opt): for t in ['train', 'valid', 'vocab']: pattern = (((opt.save_data + '.') + t) + '*.pt') if glob.glob(pattern): sys.stderr.write(('Please backup exisiting pt file: %s, to avoid tampering!\n' % pattern)) sys.exit(1)
class BaseDHDataset(BaseDataset): def __init__(self, pipeline, test_mode=False): super().__init__(pipeline, test_mode) def scan_folder(path): if isinstance(path, (str, Path)): path = str(path) else: raise TypeError(f"'path' must be a str or a Path object, but received {type(path)}.") images = list(scandir(path, suffix=IMG_EXTENSIONS, recursive=False)) images = [osp.join(path, v) for v in images] images.sort() assert images, f'{path} has no valid image file.' return images def __getitem__(self, idx): results = copy.deepcopy(self.data_infos[idx]) results['scale'] = 1 return self.pipeline(results) def evaluate(self, results, logger=None): if (not isinstance(results, list)): raise TypeError(f'results must be a list, but got {type(results)}') assert (len(results) == len(self)), f'The length of results is not equal to the dataset len: {len(results)} != {len(self)}' results = [res['eval_result'] for res in results] eval_result = defaultdict(list) for res in results: for (metric, val) in res.items(): eval_result[metric].append(val) for (metric, val_list) in eval_result.items(): assert (len(val_list) == len(self)), f'Length of evaluation result of {metric} is {len(val_list)}, should be {len(self)}' eval_result = {metric: (sum(values) / len(self)) for (metric, values) in eval_result.items()} return eval_result
def sort_imports(file, check_only=True): with open(file, encoding='utf-8') as f: code = f.read() if ('_import_structure' not in code): return main_blocks = split_code_in_indented_blocks(code, start_prompt='_import_structure = {', end_prompt='if TYPE_CHECKING:') for block_idx in range(1, (len(main_blocks) - 1)): block = main_blocks[block_idx] block_lines = block.split('\n') line_idx = 0 while ((line_idx < len(block_lines)) and ('_import_structure' not in block_lines[line_idx])): if ('import dummy' in block_lines[line_idx]): line_idx = len(block_lines) else: line_idx += 1 if (line_idx >= len(block_lines)): continue internal_block_code = '\n'.join(block_lines[line_idx:(- 1)]) indent = get_indent(block_lines[1]) internal_blocks = split_code_in_indented_blocks(internal_block_code, indent_level=indent) pattern = (_re_direct_key if ('_import_structure = {' in block_lines[0]) else _re_indirect_key) keys = [(pattern.search(b).groups()[0] if (pattern.search(b) is not None) else None) for b in internal_blocks] keys_to_sort = [(i, key) for (i, key) in enumerate(keys) if (key is not None)] sorted_indices = [x[0] for x in sorted(keys_to_sort, key=(lambda x: x[1]))] count = 0 reorderded_blocks = [] for i in range(len(internal_blocks)): if (keys[i] is None): reorderded_blocks.append(internal_blocks[i]) else: block = sort_objects_in_import(internal_blocks[sorted_indices[count]]) reorderded_blocks.append(block) count += 1 main_blocks[block_idx] = '\n'.join(((block_lines[:line_idx] + reorderded_blocks) + [block_lines[(- 1)]])) if (code != '\n'.join(main_blocks)): if check_only: return True else: print(f'Overwriting {file}.') with open(file, 'w', encoding='utf-8') as f: f.write('\n'.join(main_blocks))
class mit_b4_kd(Segformer): def __init__(self, **kwargs): super(mit_b4_kd, self).__init__(patch_size=4, embed_dims=[64, 128, 320, 512], num_heads=[1, 2, 5, 8], mlp_ratios=[4, 4, 4, 4], qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-06), depths=[3, 8, 27, 3], sr_ratios=[8, 4, 2, 1], drop_rate=0.0, drop_path_rate=0.1, decoder_dim=768)
def test_visualize_transforms_n_2(download_functional_test_files): __data_testing_dir__ = Path(__tmp_dir__, 'data_functional_testing') __input_file__ = Path(__data_testing_dir__, 'sub-unf01/anat/sub-unf01_T1w.nii.gz') __output_dir__ = Path(__tmp_dir__, 'output_visualize_transforms_n_2') __config_file__ = Path(__data_testing_dir__, 'model_config.json') __label_file__ = Path(__data_testing_dir__, 'derivatives/labels/sub-test001/anat/sub-unf01_T1w_seg-manual.nii.gz') visualize_transforms.main(args=['--input', str(__input_file__), '--output', str(__output_dir__), '--config', str(__config_file__), '-r', str(__label_file__), '-n', '2']) assert __output_dir__.exists() output_files = [f.name for f in __output_dir__.iterdir()] assert (len(output_files) == 10) for output_file in output_files: assert ('Resample' in output_file) assert ('slice' in output_file) assert ('.png' in output_file)
def plot_clep_median_error(metric, all_errors): median_error = [] for td in range(1, 22): median_error.append(np.median(all_errors['ensemble'][td][:71])) plt.figure(figsize=(4, 3), dpi=200) ax = plt.subplot(111) plt.plot(np.arange(1, 22, 1), median_error, color='lightcoral', linewidth=3) if (metric == 'mae'): plt.ylabel('Median of raw scale MAE', fontsize=15) elif (metric == 'mape'): plt.ylabel('Median of MAPE', fontsize=15) elif (metric == 'sqrt'): plt.ylabel('Median of square root MAE', fontsize=15) plt.xlabel('Horizon', fontsize=15) plt.yticks(fontsize=12) plt.xticks(fontsize=12) ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) plt.xticks([1, 6, 11, 16, 21]) plt.tight_layout() filename = os.path.join(result_dir, f'median_{metric}_clep_21_day.pdf') plt.savefig(filename)
def main(): parser = ArgumentParser('Transformers CLI tool', usage='transformers-cli <command> [<args>]') commands_parser = parser.add_subparsers(help='transformers-cli command helpers') ConvertCommand.register_subcommand(commands_parser) DownloadCommand.register_subcommand(commands_parser) EnvironmentCommand.register_subcommand(commands_parser) RunCommand.register_subcommand(commands_parser) ServeCommand.register_subcommand(commands_parser) UserCommands.register_subcommand(commands_parser) AddNewModelCommand.register_subcommand(commands_parser) AddNewModelLikeCommand.register_subcommand(commands_parser) LfsCommands.register_subcommand(commands_parser) PTtoTFCommand.register_subcommand(commands_parser) args = parser.parse_args() if (not hasattr(args, 'func')): parser.print_help() exit(1) service = args.func(args) service.run()
def loss_function(real, pred, loss_object): loss_ = loss_object(real, pred) return tf.reduce_mean(loss_)
def KL_Divergence(summary_freq, doc_freq): sum_val = 0 for (w, f) in summary_freq.items(): if (w in doc_freq): sum_val += (f * math.log((f / float(doc_freq[w])))) if np.isnan(sum_val): raise Exception('KL_Divergence returns NaN') return sum_val
def switchInterface(screen, cfg): screen.fill(cfg.BACKGROUNDCOLOR) clock = pygame.time.Clock() while True: button_1 = Button(screen, (95, 150), 'next', cfg) button_2 = Button(screen, (95, 305), 'exit', cfg) for event in pygame.event.get(): if (event.type == pygame.QUIT): pygame.quit() sys.exit() if (event.type == pygame.MOUSEBUTTONDOWN): if button_1.collidepoint(pygame.mouse.get_pos()): return elif button_2.collidepoint(pygame.mouse.get_pos()): pygame.quit() sys.exit(0) clock.tick(60) pygame.display.update()
def get_activation(model, activation, name): def hook(model, input, output): activation[name] = output.cpu().detach() return hook
class BigdlNativeLLM(LLM): logging.warning('BigdlNativeLLM has been deprecated, please switch to the new LLM API for sepcific models.') model_family: str = 'llama' family_info = {'llama': {'module': 'bigdl.llm.models', 'class': 'Llama'}, 'bloom': {'module': 'bigdl.llm.models', 'class': 'Bloom'}, 'gptneox': {'module': 'bigdl.llm.models', 'class': 'Gptneox'}, 'starcoder': {'module': 'bigdl.llm.models', 'class': 'Starcoder'}, 'chatglm': {'module': 'bigdl.llm.ggml.model.chatglm', 'class': 'ChatGLM'}} client: Any model_path: str lora_base: Optional[str] = None lora_path: Optional[str] = None n_ctx: int = Field(512, alias='n_ctx') n_parts: int = Field((- 1), alias='n_parts') seed: int = Field((- 1), alias='seed') f16_kv: bool = Field(True, alias='f16_kv') logits_all: bool = Field(False, alias='logits_all') vocab_only: bool = Field(False, alias='vocab_only') use_mlock: bool = Field(False, alias='use_mlock') n_threads: Optional[int] = Field((- 1), alias='n_threads') n_batch: Optional[int] = Field(512, alias='n_batch') n_gpu_layers: Optional[int] = Field(0, alias='n_gpu_layers') suffix: Optional[str] = Field(None) max_tokens: Optional[int] = 256 temperature: Optional[float] = 0.8 top_p: Optional[float] = 0.95 logprobs: Optional[int] = Field(None) echo: Optional[bool] = False stop: Optional[List[str]] = [] repeat_penalty: Optional[float] = 1.1 top_k: Optional[int] = 40 last_n_tokens_size: Optional[int] = 64 use_mmap: Optional[bool] = True streaming: bool = True _validator() def validate_environment(cls, values: Dict) -> Dict: model_path = values['model_path'] model_param_names = ['lora_path', 'lora_base', 'n_ctx', 'n_parts', 'seed', 'f16_kv', 'logits_all', 'vocab_only', 'use_mlock', 'n_threads', 'n_batch', 'use_mmap', 'last_n_tokens_size'] model_params = {k: values[k] for k in model_param_names} if (values['n_gpu_layers'] is not None): model_params['n_gpu_layers'] = values['n_gpu_layers'] model_family = values['model_family'].lower() if (model_family not in list(values['family_info'].keys())): raise ValueError(("Model family '%s' is not supported. Valid values are %s" % (values['model_family'], ','.join(list(values['family_info'].keys()))))) try: b_info = values['family_info'][model_family] module = importlib.import_module(b_info['module']) class_ = getattr(module, b_info['class']) values['client'] = class_(model_path, **model_params) except ImportError: raise ModuleNotFoundError('Could not import bigdl-llm library. Please install the bigdl-llm library to use this embedding model: pip install bigdl-llm') except Exception as e: raise ValueError(f'Could not load model from path: {model_path}. Please make sure the model family {model_family} matches the model you want to load.Received error {e}') return values def _default_params(self) -> Dict[(str, Any)]: return {'suffix': self.suffix, 'max_tokens': self.max_tokens, 'temperature': self.temperature, 'top_p': self.top_p, 'logprobs': self.logprobs, 'echo': self.echo, 'stop_sequences': self.stop, 'repeat_penalty': self.repeat_penalty, 'top_k': self.top_k} def _identifying_params(self) -> Dict[(str, Any)]: return {**{'model_path': self.model_path, 'model_family': self.model_family}, **self._default_params} def _llm_type(self) -> str: return 'BigDL' def _get_parameters(self, stop: Optional[List[str]]=None) -> Dict[(str, Any)]: if (self.stop and (stop is not None)): raise ValueError('`stop` found in both the input and default params.') params = self._default_params params.pop('stop_sequences') params['stop'] = (self.stop or stop or []) return params def _call(self, prompt: str, stop: Optional[List[str]]=None, run_manager: Optional[CallbackManagerForLLMRun]=None) -> str: if self.streaming: combined_text_output = '' for token in self.stream(prompt=prompt, stop=stop, run_manager=run_manager): combined_text_output += token['choices'][0]['text'] return combined_text_output else: params = self._get_parameters(stop) result = self.client(prompt=prompt, **params) return result['choices'][0]['text'] def stream(self, prompt: str, stop: Optional[List[str]]=None, run_manager: Optional[CallbackManagerForLLMRun]=None) -> Generator[(Dict, None, None)]: params = self._get_parameters(stop) result = self.client(prompt=prompt, stream=True, **params) for chunk in result: token = chunk['choices'][0]['text'] log_probs = chunk['choices'][0].get('logprobs', None) if run_manager: run_manager.on_llm_new_token(token=token, verbose=self.verbose, log_probs=log_probs) (yield chunk) def get_num_tokens(self, text: str) -> int: tokenized_text = self.client.tokenize(text.encode('utf-8')) return len(tokenized_text)
def rollouts_to_dataframe(rollouts: Iterable[Rollout], avg_over_repeats: bool=True, index_value_precision: Optional[int]=None) -> pd.DataFrame: rollouts = [(rollout.rollout_params, rollout.metrics) for rollout in rollouts] index_cols = list(rollouts[0][0].keys()) df = pd.DataFrame([{**params, **metrics} for (params, metrics) in rollouts]) if (index_value_precision is not None): for col in index_cols: df[col] = df[col].round(index_value_precision).astype(str) if (len(index_cols) > 0): if avg_over_repeats: df = df.groupby(index_cols).mean().reset_index() df = df.set_index(index_cols) return df
def fastfood_torched(x, DD, param_list=None, device=0): dd = x.size(0) if (not param_list): (BB, Pi, GG, divisor, LL) = fastfood_vars(DD, device=device) else: (BB, Pi, GG, divisor, LL) = param_list dd_pad = F.pad(x, pad=(0, (LL - dd)), value=0, mode='constant') mul_1 = torch.mul(BB, dd_pad) mul_2 = fast_walsh_hadamard_torched(mul_1, 0, normalize=False) mul_3 = mul_2[Pi] mul_4 = torch.mul(mul_3, GG) mul_5 = fast_walsh_hadamard_torched(mul_4, 0, normalize=False) ret = torch.div(mul_5[:DD], (divisor * np.sqrt((float(DD) / LL)))) return ret
class BasicConv(nn.Module): def __init__(self, in_channels, out_channels, deconv=False, is_3d=False, bn=True, relu=True, **kwargs): super(BasicConv, self).__init__() self.relu = relu self.use_bn = bn if is_3d: if deconv: self.conv = nn.ConvTranspose3d(in_channels, out_channels, bias=False, **kwargs) else: self.conv = nn.Conv3d(in_channels, out_channels, bias=False, **kwargs) self.bn = nn.BatchNorm3d(out_channels) else: if deconv: self.conv = nn.ConvTranspose2d(in_channels, out_channels, bias=False, **kwargs) else: self.conv = nn.Conv2d(in_channels, out_channels, bias=False, **kwargs) self.bn = nn.BatchNorm2d(out_channels) self.LeakyReLU = nn.LeakyReLU() def forward(self, x): x = self.conv(x) if self.use_bn: x = self.bn(x) if self.relu: x = self.LeakyReLU(x) return x
def read_data(filenames, is_test=False): instances = [] done = set() for filename in filenames: name = filename for ending in ['.annotation.txt', '.ascii.txt', '.raw.txt', '.tok.txt']: if filename.endswith(ending): name = filename[:(- len(ending))] if (name in done): continue done.add(name) text_ascii = [l.strip().split() for l in open((name + '.ascii.txt'))] text_tok = [] for l in open((name + '.tok.txt')): l = l.strip().split() if ((len(l) > 0) and (l[(- 1)] == '</s>')): l = l[:(- 1)] if ((len(l) == 0) or (l[0] != '<s>')): l.insert(0, '<s>') text_tok.append(l) (info, target_info) = lines_to_info(text_ascii) links = {} if is_test: for i in range(args.test_start, min(args.test_end, len(text_ascii))): links[i] = [] else: for line in open((name + '.annotation.txt')): nums = [int(v) for v in line.strip().split() if (v != '-')] links.setdefault(max(nums), []).append(min(nums)) for (link, nums) in links.items(): instances.append(((name + '.annotation.txt'), link, nums, text_ascii, text_tok, info, target_info)) return instances
def get_tensorflow_node_shape_attr(node): dims = None try: shape = get_tensorflow_node_attr(node, 'shape') if (not shape.unknown_rank): dims = [int(d.size) for d in shape.dim] except: pass return dims
def alpha_repair_compilation_rates() -> CompilationRateResult: compile_rates_1000: list[float] = [] compile_rates_100: list[float] = [] compile_rates_30: list[float] = [] compile_rates_5000: list[float] = [] for file in Path('data/large/uniapr_pfl').glob('*result.txt'): assert any((file.name.startswith(bug_id) for bug_id in rq_utils.D4J1_CONSIDERED_BUGS)) t = file.read_text() patches = t.split('') top_100_compile = 0 top_30_compile = 0 top_1000_compile = 0 compiled = 0 total = 0 for patch in patches: patch_number = 0 if ('Patch Number :' in patch): for line in patch.splitlines(): if ('Patch Number :' in line): patch_number = int(line.split('Patch Number :')[(- 1)].strip()) if (('Compiled Error' not in patch) and ('Syntax Error' not in patch)): assert (patch_number != 0) if (patch_number <= 100): top_100_compile += 1 if (patch_number <= 30): top_30_compile += 1 if (patch_number <= 1000): top_1000_compile += 1 compiled += 1 if (patch_number > 5000): break total += 1 if (total == 0): continue compile_rates_5000.append((compiled / total)) compile_rates_1000.append((top_1000_compile / 1000)) compile_rates_30.append((top_30_compile / 30)) compile_rates_100.append((top_100_compile / 100)) return (f'{(st.mean(compile_rates_30) * 100):.0f}%', f'{(st.mean(compile_rates_100) * 100):.0f}%', f'{(st.mean(compile_rates_1000) * 100):.0f}%', f'{(st.mean(compile_rates_5000) * 100):.0f}%')
def create_path(node1, node2, corridor_radius, map): coin_flip = np.random.random() (x1, x2) = sorted([node1[0], node2[0]]) (y1, y2) = sorted([node1[1], node2[1]]) if (get_constellation(node1, node2) == 1): if (coin_flip >= 0.5): map[(slice((x1 - corridor_radius), ((x1 + corridor_radius) + 1)), range((y1 - corridor_radius), ((y2 + 1) + corridor_radius), 1))] = 0 map[(range((x1 - corridor_radius), ((x2 + 1) + corridor_radius), 1), slice((y1 - corridor_radius), ((y1 + corridor_radius) + 1)))] = 0 else: map[(slice((x2 - corridor_radius), ((x2 + corridor_radius) + 1)), range((y1 - corridor_radius), ((y2 + 1) + corridor_radius), 1))] = 0 map[(range((x1 - corridor_radius), ((x2 + 1) + corridor_radius), 1), slice((y2 - corridor_radius), ((y2 + corridor_radius) + 1)))] = 0 elif (coin_flip >= 0.5): map[(slice((x1 - corridor_radius), ((x1 + corridor_radius) + 1)), range((y1 - corridor_radius), ((y2 + 1) + corridor_radius), 1))] = 0 map[(range((x1 - corridor_radius), ((x2 + 1) + corridor_radius), 1), slice((y2 - corridor_radius), ((y2 + corridor_radius) + 1)))] = 0 else: map[(slice((x2 - corridor_radius), ((x2 + corridor_radius) + 1)), range((y1 - corridor_radius), ((y2 + 1) + corridor_radius), 1))] = 0 map[(range((x1 - corridor_radius), ((x2 + 1) + corridor_radius), 1), slice((y1 - corridor_radius), ((y1 + corridor_radius) + 1)))] = 0
def import_fns(path, file, fns_name='StaticFns'): full_path = os.path.join(path, file) import_path = full_path.replace('/', '.') module = importlib.import_module(import_path) fns = getattr(module, fns_name) return fns
class TuckER(torch.nn.Module): def __init__(self, d, d1, d2, **kwargs): super(TuckER, self).__init__() self.E = torch.nn.Embedding(len(d.entities), d1, padding_idx=0) self.R = torch.nn.Embedding(len(d.relations), d2, padding_idx=0) self.W = torch.nn.Parameter(torch.tensor(np.random.uniform((- 1), 1, (d2, d1, d1)), dtype=torch.float, device='cuda', requires_grad=True)) self.input_dropout = torch.nn.Dropout(kwargs['input_dropout']) self.hidden_dropout1 = torch.nn.Dropout(kwargs['hidden_dropout1']) self.hidden_dropout2 = torch.nn.Dropout(kwargs['hidden_dropout2']) self.loss = torch.nn.BCELoss() self.bn0 = torch.nn.BatchNorm1d(d1) self.bn1 = torch.nn.BatchNorm1d(d1) def init(self): xavier_normal_(self.E.weight.data) xavier_normal_(self.R.weight.data) def forward(self, e1_idx, r_idx): e1 = self.E(e1_idx) x = self.bn0(e1) x = self.input_dropout(x) x = x.view((- 1), 1, e1.size(1)) r = self.R(r_idx) W_mat = torch.mm(r, self.W.view(r.size(1), (- 1))) W_mat = W_mat.view((- 1), e1.size(1), e1.size(1)) W_mat = self.hidden_dropout1(W_mat) x = torch.bmm(x, W_mat) x = x.view((- 1), e1.size(1)) x = self.bn1(x) x = self.hidden_dropout2(x) x = torch.mm(x, self.E.weight.transpose(1, 0)) pred = torch.sigmoid(x) return pred
_func def ClearAllIntegers(data): if (type(data) == int): return 0 if (type(data) == list): for i in range(0, len(data)): data[i] = ClearAllIntegers(data[i]) if (type(data) == dict): for (k, v) in data: data[k] = ClearAllIntegers(v) return data
class _PostSampledEmdAutoSavingManager(_EmdAutoSavingManager): def path(self): return get_eval_path('emd', 'postsampled', str(self._n_samples), ('%s.json' % self._model_id)) def get_inferred_cloud_dataset(self): return clouds.get_inferred_cloud_dataset(pre_sampled=False, model_id=self._model_id, n_samples=self._n_samples, **self._kwargs)
def compute_nas_score(gpu, model, resolution, batch_size): model.train() model.requires_grad_(True) model.zero_grad() if (gpu is not None): torch.cuda.set_device(gpu) model = model.cuda(gpu) network_weight_gaussian_init(model) input = torch.randn(size=[batch_size, 3, resolution, resolution]) if (gpu is not None): input = input.cuda(gpu) output = model(input) num_classes = output.shape[1] y = torch.randint(low=0, high=num_classes, size=[batch_size]) one_hot_y = F.one_hot(y, num_classes).float() if (gpu is not None): one_hot_y = one_hot_y.cuda(gpu) loss = cross_entropy(output, one_hot_y) loss.backward() norm2_sum = 0 with torch.no_grad(): for p in model.parameters(): if (hasattr(p, 'grad') and (p.grad is not None)): norm2_sum += (torch.norm(p.grad) ** 2) grad_norm = float(torch.sqrt(norm2_sum)) return grad_norm
class LayoutLMv3FeatureExtractor(LayoutLMv3ImageProcessor): def __init__(self, *args, **kwargs) -> None: warnings.warn('The class LayoutLMv3FeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please use LayoutLMv3ImageProcessor instead.', FutureWarning) super().__init__(*args, **kwargs)
class DataArguments(): corpus_path: str = field() out_corpus_dir: str = field() query_path: str = field() out_query_dir: str = field() qrel_path: str = field(default=None)
def postprocess(B, graph_thres=0.3): B = np.copy(B) B[(np.abs(B) <= graph_thres)] = 0 (B, _) = threshold_till_dag(B) return B
def parse_log(root: Path, name: str='') -> Dict: with root.joinpath('log.txt').open() as f: result = {'path': root, 'name': (f'({name}) {root.name}' if name else root.name), 'params': {}, 'train': [], 'valid': []} for line in f: param_match = re.match('^\\s+- (\\w+) : (.+)', line) if param_match: (key, value) = param_match.groups() result['params'][key] = value elif line.startswith('| epoch'): item = {field: float(re.search(f'{field}\s+(\d+\.?\d*)', line).groups()[0]) for field in ['step', 'epoch', 'lr', 'loss', 'ppl']} result['train'].append(item) elif line.startswith('| Eval'): item = {field: float(re.search(f'{field}\s+(\d+\.?\d*)', line).groups()[0]) for field in ['Eval', 'step', 'valid loss', 'valid ppl']} result['valid'].append(item) return result
def generate_problem_data(P, q, A, l, u, problem_name, sols_data={}): n = P.shape[0] m = A.shape[0] f = open((problem_name + '/data.h'), 'w') f.write((('#ifndef ' + problem_name.upper()) + '_DATA_H\n')) f.write((('#define ' + problem_name.upper()) + '_DATA_H\n')) f.write('#include "osqp.h"\n') f.write('\n\n') f.write('/* create additional data and solutions structure */\n') f.write('typedef struct {\n') for (key, value) in sols_data.items(): if isinstance(value, str): f.write(('c_int %s;\n' % key)) elif isinstance(value, np.ndarray): if isinstance(value.flatten(order='F')[0], int): f.write(('c_int * %s;\n' % key)) elif isinstance(value.flatten(order='F')[0], float): f.write(('c_float * %s;\n' % key)) elif isinstance(value, int): f.write(('c_int %s;\n' % key)) elif isinstance(value, float): f.write(('c_float %s;\n' % key)) f.write(('} %s_sols_data;\n\n' % problem_name)) f.write('/* function prototypes */\n') f.write(('OSQPData * generate_problem_%s();\n' % problem_name)) f.write(('void clean_problem_%s(OSQPData * data);\n' % problem_name)) f.write(('%s_sols_data * generate_problem_%s_sols_data();\n' % (problem_name, problem_name))) f.write(('void clean_problem_%s_sols_data(%s_sols_data * data);\n' % (problem_name, problem_name))) f.write('\n\n') f.write('/* function to generate QP problem data */\n') f.write(('OSQPData * generate_problem_%s(){\n\n' % problem_name)) f.write('OSQPData * data = (OSQPData *)c_malloc(sizeof(OSQPData));\n\n') f.write('// Problem dimensions\n') write_int(f, n, 'n', 'data') write_int(f, m, 'm', 'data') f.write('\n') f.write('// Problem vectors\n') write_vec_float(f, l, 'l', 'data') write_vec_float(f, u, 'u', 'data') write_vec_float(f, q, 'q', 'data') f.write('\n') write_mat_sparse(f, A, 'A', 'data') write_mat_sparse(f, P, 'P', 'data') f.write('return data;\n\n') f.write('}\n\n') f.write('/* function to clean problem data structure */\n') f.write(('void clean_problem_%s(OSQPData * data){\n\n' % problem_name)) f.write('// Clean vectors\n') clean_vec(f, 'l', 'data') clean_vec(f, 'u', 'data') clean_vec(f, 'q', 'data') f.write('\n') f.write('//Clean Matrices\n') clean_mat(f, 'A', 'data') clean_mat(f, 'P', 'data') f.write('\n') f.write('c_free(data);\n\n') f.write('}\n\n') f.write('/* function to define solutions and additional data struct */\n') f.write(('%s_sols_data * generate_problem_%s_sols_data(){\n\n' % (problem_name, problem_name))) f.write(('%s_sols_data * data = (%s_sols_data *)c_malloc(sizeof(%s_sols_data));\n\n' % (problem_name, problem_name, problem_name))) for (key, value) in sols_data.items(): if isinstance(value, str): if (value == 'optimal'): f.write(('data->%s = %s;\n' % (key, 'OSQP_SOLVED'))) elif (value == 'optimal_inaccurate'): f.write(('data->%s = %s;\n' % (key, 'OSQP_SOLVED_INACCURATE'))) elif (value == 'primal_infeasible'): f.write(('data->%s = %s;\n' % (key, 'OSQP_PRIMAL_INFEASIBLE'))) elif (value == 'primal_infeasible_inaccurate'): f.write(('data->%s = %s;\n' % (key, 'OSQP_PRIMAL_INFEASIBLE_INACCURATE'))) elif (value == 'dual_infeasible'): f.write(('data->%s = %s;\n' % (key, 'OSQP_DUAL_INFEASIBLE'))) elif (value == 'dual_infeasible_inaccurate'): f.write(('data->%s = %s;\n' % (key, 'OSQP_DUAL_INFEASIBLE_INACCURATE'))) if (type(value) is np.ndarray): if isinstance(value.flatten(order='F')[0], int): write_vec_int(f, value.flatten(order='F'), key, 'data') elif isinstance(value.flatten(order='F')[0], float): write_vec_float(f, value.flatten(order='F'), key, 'data') elif isinstance(value, int): write_int(f, value, key, 'data') elif isinstance(value, float): write_float(f, value, key, 'data') f.write('\nreturn data;\n\n') f.write('}\n\n') f.write('/* function to clean solutions and additional data struct */\n') f.write(('void clean_problem_%s_sols_data(%s_sols_data * data){\n\n' % (problem_name, problem_name))) for (key, value) in sols_data.items(): if (type(value) is np.ndarray): clean_vec(f, key, 'data') f.write('\nc_free(data);\n\n') f.write('}\n\n') f.write('#endif\n') f.close()
class FTB_block(nn.Module): def __init__(self, dim_in, dim_out): super().__init__() self.dim_in = dim_in self.dim_out = dim_out self.conv1 = nn.Conv2d(self.dim_in, self.dim_out, 1, stride=1, padding=0, bias=False) self.conv2 = nn.Conv2d(self.dim_out, self.dim_out, 3, stride=1, padding=2, dilation=2, bias=True) self.bn1 = nn.BatchNorm2d(self.dim_out, momentum=0.5) self.relu = nn.ReLU(inplace=True) self.conv3 = nn.Conv2d(self.dim_out, self.dim_out, 3, stride=1, padding=2, dilation=2, bias=False) def forward(self, x): x = self.conv1(x) residual = x out = self.conv2(x) out = self.bn1(out) out = self.relu(out) out = self.conv3(out) out += residual out = self.relu(out) return out
def get_memory(tn_orig, rpn): tn = tn_orig.clone() cost = [] for item in rpn: if (item == (- 1)): c1 = cost.pop() c0 = cost.pop() index1 = c1[0] index0 = c0[0] t0 = tn.tensors[index0] t1 = tn.tensors[index1] (bc, br0, br1) = tn.find_bonds(index0, index1) mem_start = (c0[2] + c1[2]) mem_end = 1.0 for b in (br0 + br1): mem_end *= BOND_DIMS[b] mem_req = max((c0[1] + c1[2]), (c0[1] + c1[3]), (c0[2] + c1[1]), (c0[3] + c1[1]), ((mem_end + c0[3]) + c1[3])) tn = tn.contract(index0, index1, bc, br0, br1) cost.append((index0, mem_req, mem_start, mem_end)) else: t = tn.tensors[item] val = 1.0 for b in t.bonds: val *= BOND_DIMS[b] cost.append((item, val, val, val)) return cost[0][1]
class Gather(Function): def forward(ctx, target_device, dim, *inputs): assert all(map((lambda i: i.is_cuda), inputs)) ctx.target_device = target_device ctx.dim = dim ctx.input_gpus = tuple(map((lambda i: i.get_device()), inputs)) ctx.input_sizes = tuple(map((lambda i: i.size(ctx.dim)), inputs)) return comm.gather(inputs, ctx.dim, ctx.target_device) def backward(ctx, grad_output): return ((None, None) + Scatter.apply(ctx.input_gpus, ctx.input_sizes, ctx.dim, grad_output))
class TensorflowFeatureExtractor(BaseFeatureExtractor): def __init__(self, device: Optional[str]=None): super().__init__(backend='tensorflow') self.device = device if isinstance(device, str): self.device = device.replace('cuda', 'gpu') def _predict(self, batch_images): if (batch_images.dtype == tf.uint8): batch_images = tf.map_fn(self.transform, batch_images, dtype=tf.float32) with (tf.device(self.device) if self.device else no_scope()): return self.model(batch_images, training=False)[0] def __call__(self, obj, **kwargs): import tensorflow as tf if isinstance(obj, sf.WSI): grid = features_from_slide(self, obj, preprocess_fn=self.transform, **kwargs) return np.ma.masked_where((grid == sf.heatmap.MASK), grid) elif kwargs: raise ValueError(f'{self.__class__.__name__} does not accept keyword arguments when extracting features from a batch of images.') assert (obj.dtype in (tf.float32, tf.uint8)) return self._predict(obj)
class SoftCrossEntropyLoss(nn.Module): __constants__ = ['reduction', 'ignore_index', 'smooth_factor'] def __init__(self, reduction: str='mean', smooth_factor: Optional[float]=None, ignore_index: Optional[int]=(- 100), dim: int=1): super().__init__() self.smooth_factor = smooth_factor self.ignore_index = ignore_index self.reduction = reduction self.dim = dim def forward(self, y_pred: torch.Tensor, y_true: torch.Tensor) -> torch.Tensor: log_prob = F.log_softmax(y_pred, dim=self.dim) return label_smoothed_nll_loss(log_prob, y_true, epsilon=self.smooth_factor, ignore_index=self.ignore_index, reduction=self.reduction, dim=self.dim)
def train(opt): (opt.use_fc, opt.use_att) = utils.if_use_feat(opt.caption_model) opt.use_att = True opt.use_fc = True if opt.use_box: opt.att_feat_size = (opt.att_feat_size + 5) loader = DataLoader(opt) opt.vocab_size = loader.vocab_size opt.seq_length = loader.seq_length tb_summary_writer = (tb and tb.SummaryWriter(opt.checkpoint_path)) infos = {} histories = {} if (opt.start_from is not None): with open(os.path.join(opt.start_from, (('infos_' + opt.id) + '.pkl')), 'rb') as f: infos = utils.pickle_load(f) saved_model_opt = infos['opt'] need_be_same = ['caption_model', 'rnn_type', 'rnn_size', 'num_layers'] for checkme in need_be_same: assert (vars(saved_model_opt)[checkme] == vars(opt)[checkme]), ("Command line argument and saved model disagree on '%s' " % checkme) if os.path.isfile(os.path.join(opt.start_from, (('histories_' + opt.id) + '.pkl'))): with open(os.path.join(opt.start_from, (('histories_' + opt.id) + '.pkl')), 'rb') as f: histories = utils.pickle_load(f) else: infos['iter'] = 0 infos['epoch'] = 0 infos['iterators'] = loader.iterators infos['split_ix'] = loader.split_ix infos['vocab'] = loader.get_vocab() infos['opt'] = opt iteration = infos.get('iter', 0) epoch = infos.get('epoch', 0) val_result_history = histories.get('val_result_history', {}) loss_history = histories.get('loss_history', {}) lr_history = histories.get('lr_history', {}) ss_prob_history = histories.get('ss_prob_history', {}) loader.iterators = infos.get('iterators', loader.iterators) loader.split_ix = infos.get('split_ix', loader.split_ix) if (opt.load_best_score == 1): best_val_score = infos.get('best_val_score', None) if (opt.teacher_caption_model != opt.caption_model): teacher_opt = copy.deepcopy(opt) teacher_opt.caption_model = opt.teacher_caption_model teacher_opt.start_from = None if (opt.teacher_caption_model == 'topdown'): teacher_opt.num_layers = 1 teacher_opt.rnn_size = 1024 if (opt.teacher_caption_model == 'show_tell'): teacher_opt.num_layers = 1 teacher_opt.rnn_size = 1024 if (opt.teacher_caption_model == 'transformer'): teacher_opt.num_layers = 6 teacher_opt.rnn_size = 2048 else: teacher_opt = opt teacher = models.setup(teacher_opt).cuda() if (opt.teacher_caption_model == 'topdown'): teacher.load_state_dict(torch.load('save/up-down-baseline/model-best.pth')) elif (opt.teacher_caption_model == 'show_tell'): teacher.load_state_dict(torch.load('save/student-showtell-teacher-updown-1.5/model-best.pth')) elif (opt.teacher_caption_model == 'transformer'): teacher.load_state_dict(torch.load('save/nsc-transformer-baseline/model-best.pth')) teacher = torch.nn.DataParallel(teacher) teacher.eval() model = models.setup(opt).cuda() if ((opt.start_from == None) and (opt.caption_model == 'sat') and (opt.K > 1)): model.load_state_dict(torch.load('save/nsc-transformer-baseline/model-best.pth')) dp_model = torch.nn.DataParallel(model) lw_model = LossWrapper(model, teacher, opt) dp_lw_model = torch.nn.DataParallel(lw_model) epoch_done = True dp_lw_model.train() if opt.noamopt: assert ((opt.caption_model == 'transformer') or (opt.caption_model == 'sat')), 'noamopt can only work with transformer' optimizer = utils.get_std_opt(model, factor=opt.noamopt_factor, warmup=opt.noamopt_warmup) optimizer._step = iteration elif opt.reduce_on_plateau: optimizer = utils.build_optimizer(model.parameters(), opt) optimizer = utils.ReduceLROnPlateau(optimizer, factor=0.5, patience=3) else: optimizer = utils.build_optimizer(model.parameters(), opt) if ((vars(opt).get('start_from', None) is not None) and os.path.isfile(os.path.join(opt.start_from, 'optimizer.pth'))): optimizer.load_state_dict(torch.load(os.path.join(opt.start_from, 'optimizer.pth'))) def save_checkpoint(model, infos, optimizer, histories=None, append=''): if (len(append) > 0): append = ('-' + append) if (not os.path.isdir(opt.checkpoint_path)): os.makedirs(opt.checkpoint_path) checkpoint_path = os.path.join(opt.checkpoint_path, ('model%s.pth' % append)) torch.save(model.state_dict(), checkpoint_path) print('model saved to {}'.format(checkpoint_path)) optimizer_path = os.path.join(opt.checkpoint_path, ('optimizer%s.pth' % append)) torch.save(optimizer.state_dict(), optimizer_path) with open(os.path.join(opt.checkpoint_path, (('infos_' + opt.id) + ('%s.pkl' % append))), 'wb') as f: utils.pickle_dump(infos, f) if histories: with open(os.path.join(opt.checkpoint_path, (('histories_' + opt.id) + ('%s.pkl' % append))), 'wb') as f: utils.pickle_dump(histories, f) try: while True: if epoch_done: if ((not opt.noamopt) and (not opt.reduce_on_plateau)): if ((epoch > opt.learning_rate_decay_start) and (opt.learning_rate_decay_start >= 0)): frac = ((epoch - opt.learning_rate_decay_start) // opt.learning_rate_decay_every) decay_factor = (opt.learning_rate_decay_rate ** frac) opt.current_lr = (opt.learning_rate * decay_factor) else: opt.current_lr = opt.learning_rate utils.set_lr(optimizer, opt.current_lr) if ((epoch > opt.scheduled_sampling_start) and (opt.scheduled_sampling_start >= 0)): frac = ((epoch - opt.scheduled_sampling_start) // opt.scheduled_sampling_increase_every) opt.ss_prob = min((opt.scheduled_sampling_increase_prob * frac), opt.scheduled_sampling_max_prob) model.ss_prob = opt.ss_prob if ((opt.self_critical_after != (- 1)) and (epoch >= opt.self_critical_after)): sc_flag = True init_scorer(opt.cached_tokens) else: sc_flag = False epoch_done = False start = time.time() data = loader.get_batch('train') print('Read data:', (time.time() - start)) torch.cuda.synchronize() start = time.time() tmp = [data['fc_feats'], data['att_feats'], data['labels'], data['masks'], data['att_masks']] tmp = [(_ if (_ is None) else _.cuda()) for _ in tmp] (fc_feats, att_feats, labels, masks, att_masks) = tmp optimizer.zero_grad() model_out = dp_lw_model(fc_feats, att_feats, labels, masks, att_masks, data['gts'], torch.arange(0, len(data['gts'])), sc_flag) loss = model_out['loss'].mean() loss.backward() utils.clip_gradient(optimizer, opt.grad_clip) optimizer.step() train_loss = loss.item() torch.cuda.synchronize() end = time.time() if (not sc_flag): print('iter {} (epoch {}), train_loss = {:.3f}, time/batch = {:.3f}'.format(iteration, epoch, train_loss, (end - start))) else: print('iter {} (epoch {}), avg_reward = {:.3f}, time/batch = {:.3f}'.format(iteration, epoch, model_out['reward'].mean(), (end - start))) iteration += 1 if data['bounds']['wrapped']: epoch += 1 epoch_done = True if ((iteration % opt.losses_log_every) == 0): add_summary_value(tb_summary_writer, 'train_loss', train_loss, iteration) if opt.noamopt: opt.current_lr = optimizer.rate() elif opt.reduce_on_plateau: opt.current_lr = optimizer.current_lr add_summary_value(tb_summary_writer, 'learning_rate', opt.current_lr, iteration) add_summary_value(tb_summary_writer, 'scheduled_sampling_prob', model.ss_prob, iteration) if sc_flag: add_summary_value(tb_summary_writer, 'avg_reward', model_out['reward'].mean(), iteration) loss_history[iteration] = (train_loss if (not sc_flag) else model_out['reward'].mean()) lr_history[iteration] = opt.current_lr ss_prob_history[iteration] = model.ss_prob infos['iter'] = iteration infos['epoch'] = epoch infos['iterators'] = loader.iterators infos['split_ix'] = loader.split_ix if ((iteration % opt.save_checkpoint_every) == 0): eval_kwargs = {'split': 'val', 'dataset': opt.input_json} eval_kwargs.update(vars(opt)) (val_loss, predictions, lang_stats) = eval_utils.eval_split(dp_model, lw_model.crit, loader, eval_kwargs) if opt.reduce_on_plateau: if ('CIDEr' in lang_stats): optimizer.scheduler_step((- lang_stats['CIDEr'])) else: optimizer.scheduler_step(val_loss) add_summary_value(tb_summary_writer, 'validation loss', val_loss, iteration) if (lang_stats is not None): for (k, v) in lang_stats.items(): add_summary_value(tb_summary_writer, k, v, iteration) val_result_history[iteration] = {'loss': val_loss, 'lang_stats': lang_stats, 'predictions': predictions} if (opt.language_eval == 1): current_score = lang_stats['CIDEr'] else: current_score = (- val_loss) best_flag = False if ((best_val_score is None) or (current_score > best_val_score)): best_val_score = current_score best_flag = True infos['best_val_score'] = best_val_score histories['val_result_history'] = val_result_history histories['loss_history'] = loss_history histories['lr_history'] = lr_history histories['ss_prob_history'] = ss_prob_history save_checkpoint(model, infos, optimizer, histories) if opt.save_history_ckpt: save_checkpoint(model, infos, optimizer, append=str(iteration)) if best_flag: save_checkpoint(model, infos, optimizer, append='best') if ((epoch >= opt.max_epochs) and (opt.max_epochs != (- 1))): break except (RuntimeError, KeyboardInterrupt): print('Save ckpt on exception ...') save_checkpoint(model, infos, optimizer) print('Save ckpt done.') stack_trace = traceback.format_exc() print(stack_trace)
class AlbuSeg3d(Repr): def __init__(self, albumentation: Callable): self.albumentation = A.ReplayCompose([albumentation]) def __call__(self, inp: np.ndarray, tar: np.ndarray): tar = tar.astype(np.uint8) input_copy = np.copy(inp) target_copy = np.copy(tar) replay_dict = self.albumentation(image=inp[0])['replay'] for (index, (input_slice, target_slice)) in enumerate(zip(inp, tar)): result = A.ReplayCompose.replay(replay_dict, image=input_slice, mask=target_slice) input_copy[index] = result['image'] target_copy[index] = result['mask'] return (input_copy, target_copy)
class TestWindowedIterator(TestBase): def test(self): for n in [0, 2, 3, 8, 9, 10, 11, 12]: seq = list(range(n)) it = WindowedIterator(NativeCheckpointableIterator(seq), 3) actual0 = list(itertools.islice(it, ((n * 3) // 10))) checkpoint = it.getstate() actual1a = list(it) it.setstate(checkpoint) actual1b = list(it) actual = (actual0 + actual1a) expected = list(zip(seq, itertools.islice(seq, 1, None), itertools.islice(seq, 2, None))) self.assertListEqual(actual, expected) self.assertListEqual(actual1a, actual1b)
class TestDatasets(unittest.TestCase): def _test_dset(self, dset, n_train, n_test, shape, output_size): dset = dset() self.assertEqual(len(dset.train_data), n_train) self.assertEqual(len(dset.test_data), n_test) self.assertEqual(dset.shape, shape) self.assertEqual(dset.train_data[[0]][0][0].shape, shape) self.assertEqual(dset.output_size, output_size) (os.getenv('TEST_DATASETS', False), 'Datasets need not be tested all the time') def test_datasets(self): datasets = [(CIFAR10, 50000, 10000, (32, 32, 3), 10), (CIFARSanityCheck, 40000, 2000, (32, 32, 3), 2), (MNIST, 60000, 10000, (28, 28, 1), 10), (partial(CanevetICML2016, N=256), (int(((256 ** 2) - ((256 ** 2) * 0.33))) + 1), int(((256 ** 2) * 0.33)), (2,), 2), (compose(partial(OntheflyAugmentedImages, augmentation_params=dict(featurewise_center=False, samplewise_center=False, featurewise_std_normalization=False, samplewise_std_normalization=False, zca_whitening=False, rotation_range=0, width_shift_range=0.1, height_shift_range=0.1, horizontal_flip=True, vertical_flip=False)), CIFAR10), 500000, 10000, (32, 32, 3), 10), (partial(PennTreeBank, 20), 887521, 70390, (20,), 10000)] for args in datasets: self._test_dset(*args) def test_zca_whitening(self): dset = ZCAWhitening(InMemoryImageDataset(np.random.rand(1000, 32, 32, 3), (np.random.rand(1000, 1) * 10).astype(np.int32), np.random.rand(1000, 32, 32, 3), (np.random.rand(1000, 1) * 10).astype(np.int32))) self.assertEqual(len(dset.train_data), 1000) self.assertEqual(len(dset.test_data), 1000) idxs = np.random.choice(len(dset.train_data), 100) (x_r, y_r) = dset.train_data[idxs] self.assertEqual(x_r.shape, (100, 32, 32, 3)) self.assertEqual(y_r.shape, (100, 10)) for i in range(10): (x, y) = dset.train_data[idxs] self.assertTrue(np.all((x_r == x))) self.assertTrue(np.all((y_r == y))) def test_image_augmentation(self): orig_dset = InMemoryImageDataset(np.random.rand(1000, 32, 32, 3), (np.random.rand(1000, 1) * 10).astype(np.int32), np.random.rand(1000, 32, 32, 3), (np.random.rand(1000, 1) * 10).astype(np.int32)) dset = OntheflyAugmentedImages(orig_dset, dict(featurewise_center=False, samplewise_center=False, featurewise_std_normalization=False, samplewise_std_normalization=False, zca_whitening=False, rotation_range=0, width_shift_range=0.1, height_shift_range=0.1, horizontal_flip=True, vertical_flip=False)) idxs = np.random.choice(len(dset.train_data), 100) (x_r, y_r) = dset.train_data[idxs] for i in range(10): (x, y) = dset.train_data[idxs] self.assertTrue(np.all((x_r == x))) self.assertTrue(np.all((y_r == y))) dset = OntheflyAugmentedImages(orig_dset, dict(featurewise_center=True, samplewise_center=False, featurewise_std_normalization=False, samplewise_std_normalization=False, zca_whitening=True, rotation_range=0, width_shift_range=0.1, height_shift_range=0.1, horizontal_flip=True, vertical_flip=False)) idxs = np.random.choice(len(dset.train_data), 100) (x_r, y_r) = dset.train_data[idxs] for i in range(10): (x, y) = dset.train_data[idxs] self.assertTrue(np.all((x_r == x))) self.assertTrue(np.all((y_r == y))) def test_generator(self): def data(): while True: (yield (np.random.rand(32, 10), np.random.rand(32, 1))) dset = GeneratorDataset(data()) self.assertEqual(dset.shape, (10,)) self.assertEqual(dset.output_size, 1) with self.assertRaises(RuntimeError): len(dset.train_data) (x, y) = dset.train_data[:10] self.assertEqual(10, len(x)) self.assertEqual(10, len(y)) (x, y) = dset.train_data[:100] self.assertEqual(100, len(x)) self.assertEqual(100, len(y)) (x, y) = dset.train_data[[1, 2, 17]] self.assertEqual(3, len(x)) self.assertEqual(3, len(y)) (x, y) = dset.train_data[99] self.assertEqual(1, len(x)) self.assertEqual(1, len(y)) with self.assertRaises(RuntimeError): len(dset.test_data) with self.assertRaises(RuntimeError): dset.test_data[0] test_data = (np.random.rand(120, 10), np.random.rand(120, 1)) dset = GeneratorDataset(data(), test_data) self.assertEqual(dset.shape, (10,)) self.assertEqual(dset.output_size, 1) with self.assertRaises(RuntimeError): len(dset.train_data) (x, y) = dset.train_data[:10] self.assertEqual(10, len(x)) self.assertEqual(10, len(y)) (x, y) = dset.train_data[:100] self.assertEqual(100, len(x)) self.assertEqual(100, len(y)) (x, y) = dset.train_data[[1, 2, 17]] self.assertEqual(3, len(x)) self.assertEqual(3, len(y)) (x, y) = dset.train_data[99] self.assertEqual(1, len(x)) self.assertEqual(1, len(y)) self.assertEqual(120, len(dset.test_data)) self.assertTrue(np.all((test_data[0] == dset.test_data[:][0]))) idxs = [10, 20, 33] self.assertTrue(np.all((test_data[1][idxs] == dset.test_data[idxs][1]))) dset = GeneratorDataset(data(), data(), 120) self.assertEqual(120, len(dset.test_data)) (x, y) = dset.test_data[:10] self.assertEqual(10, len(x)) self.assertEqual(10, len(y)) (x, y) = dset.test_data[:100] self.assertEqual(100, len(x)) self.assertEqual(100, len(y)) (x, y) = dset.test_data[[1, 2, 17]] self.assertEqual(3, len(x)) self.assertEqual(3, len(y)) (x, y) = dset.test_data[99] self.assertEqual(1, len(x)) self.assertEqual(1, len(y))
_cache() def setup_logger(name='log', logging_level='debug') -> logging.Logger: if (logging_level == 'debug'): logging_level = logging.DEBUG elif (logging_level == 'info'): logging_level = logging.INFO logger = logging.getLogger(name) logger.setLevel(logging_level) logger.propagate = False formatter = logging.Formatter('[%(asctime)s] %(name)s %(levelname)s: %(message)s', datefmt='%m/%d %H:%M:%S') ch = logging.StreamHandler(stream=sys.stdout) ch.setLevel(logging_level) ch.setFormatter(formatter) logger.addHandler(ch) return logger
def main(exp, args, num_gpu): if (args.seed is not None): random.seed(args.seed) torch.manual_seed(args.seed) cudnn.deterministic = True warnings.warn('You have chosen to seed testing. This will turn on the CUDNN deterministic setting, ') is_distributed = (num_gpu > 1) cudnn.benchmark = True rank = args.local_rank file_name = os.path.join(exp.output_dir, args.experiment_name) if (rank == 0): os.makedirs(file_name, exist_ok=True) results_folder = os.path.join(file_name, 'track_results_sort') os.makedirs(results_folder, exist_ok=True) setup_logger(file_name, distributed_rank=rank, filename='val_log.txt', mode='a') logger.info('Args: {}'.format(args)) if (args.conf is not None): exp.test_conf = args.conf if (args.nms is not None): exp.nmsthre = args.nms if (args.tsize is not None): exp.test_size = (args.tsize, args.tsize) model = exp.get_model() logger.info('Model Summary: {}'.format(get_model_info(model, exp.test_size))) val_loader = exp.get_eval_loader(args.batch_size, is_distributed, args.test) evaluator = MOTEvaluator(args=args, dataloader=val_loader, img_size=exp.test_size, confthre=exp.test_conf, nmsthre=exp.nmsthre, num_classes=exp.num_classes) torch.cuda.set_device(rank) model.cuda(rank) model.eval() if ((not args.speed) and (not args.trt)): if (args.ckpt is None): ckpt_file = os.path.join(file_name, 'best_ckpt.pth.tar') else: ckpt_file = args.ckpt logger.info('loading checkpoint') loc = 'cuda:{}'.format(rank) ckpt = torch.load(ckpt_file, map_location=loc) model.load_state_dict(ckpt['model']) logger.info('loaded checkpoint done.') if is_distributed: model = DDP(model, device_ids=[rank]) if args.fuse: logger.info('\tFusing model...') model = fuse_model(model) if args.trt: assert ((not args.fuse) and (not is_distributed) and (args.batch_size == 1)), 'TensorRT model is not support model fusing and distributed inferencing!' trt_file = os.path.join(file_name, 'model_trt.pth') assert os.path.exists(trt_file), 'TensorRT model is not found!\n Run tools/trt.py first!' model.head.decode_in_inference = False decoder = model.head.decode_outputs else: trt_file = None decoder = None (*_, summary) = evaluator.evaluate_sort(model, is_distributed, args.fp16, trt_file, decoder, exp.test_size, results_folder) logger.info(('\n' + summary)) mm.lap.default_solver = 'lap' gt_type = '_val_half' print('gt_type', gt_type) gtfiles = glob.glob(os.path.join('datasets/mot/train', '*/gt/gt{}.txt'.format(gt_type))) print('gt_files', gtfiles) tsfiles = [f for f in glob.glob(os.path.join(results_folder, '*.txt')) if (not os.path.basename(f).startswith('eval'))] logger.info('Found {} groundtruths and {} test files.'.format(len(gtfiles), len(tsfiles))) logger.info('Available LAP solvers {}'.format(mm.lap.available_solvers)) logger.info("Default LAP solver '{}'".format(mm.lap.default_solver)) logger.info('Loading files.') gt = OrderedDict([(Path(f).parts[(- 3)], mm.io.loadtxt(f, fmt='mot15-2D', min_confidence=1)) for f in gtfiles]) ts = OrderedDict([(os.path.splitext(Path(f).parts[(- 1)])[0], mm.io.loadtxt(f, fmt='mot15-2D', min_confidence=(- 1))) for f in tsfiles]) mh = mm.metrics.create() (accs, names) = compare_dataframes(gt, ts) logger.info('Running metrics') metrics = ['recall', 'precision', 'num_unique_objects', 'mostly_tracked', 'partially_tracked', 'mostly_lost', 'num_false_positives', 'num_misses', 'num_switches', 'num_fragmentations', 'mota', 'motp', 'num_objects'] summary = mh.compute_many(accs, names=names, metrics=metrics, generate_overall=True) div_dict = {'num_objects': ['num_false_positives', 'num_misses', 'num_switches', 'num_fragmentations'], 'num_unique_objects': ['mostly_tracked', 'partially_tracked', 'mostly_lost']} for divisor in div_dict: for divided in div_dict[divisor]: summary[divided] = (summary[divided] / summary[divisor]) fmt = mh.formatters change_fmt_list = ['num_false_positives', 'num_misses', 'num_switches', 'num_fragmentations', 'mostly_tracked', 'partially_tracked', 'mostly_lost'] for k in change_fmt_list: fmt[k] = fmt['mota'] print(mm.io.render_summary(summary, formatters=fmt, namemap=mm.io.motchallenge_metric_names)) metrics = (mm.metrics.motchallenge_metrics + ['num_objects']) summary = mh.compute_many(accs, names=names, metrics=metrics, generate_overall=True) print(mm.io.render_summary(summary, formatters=mh.formatters, namemap=mm.io.motchallenge_metric_names)) logger.info('Completed')