diff --git a/examples/contrib/run_swag.py b/examples/contrib/run_swag.py index a996a2ac9d..70dcca3b0b 100644 --- a/examples/contrib/run_swag.py +++ b/examples/contrib/run_swag.py @@ -76,7 +76,7 @@ class SwagExample(object): return self.__repr__() def __repr__(self): - l = [ + attributes = [ "swag_id: {}".format(self.swag_id), "context_sentence: {}".format(self.context_sentence), "start_ending: {}".format(self.start_ending), @@ -87,9 +87,9 @@ class SwagExample(object): ] if self.label is not None: - l.append("label: {}".format(self.label)) + attributes.append("label: {}".format(self.label)) - return ", ".join(l) + return ", ".join(attributes) class InputFeatures(object): diff --git a/templates/adding_a_new_model/modeling_xxx.py b/templates/adding_a_new_model/modeling_xxx.py index 5a5f76b3d0..0779b65214 100644 --- a/templates/adding_a_new_model/modeling_xxx.py +++ b/templates/adding_a_new_model/modeling_xxx.py @@ -89,25 +89,25 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path): pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): - l = re.split(r"_(\d+)", m_name) + scope_names = re.split(r"_(\d+)", m_name) else: - l = [m_name] - if l[0] == "kernel" or l[0] == "gamma": + scope_names = [m_name] + if scope_names[0] == "kernel" or scope_names[0] == "gamma": pointer = getattr(pointer, "weight") - elif l[0] == "output_bias" or l[0] == "beta": + elif scope_names[0] == "output_bias" or scope_names[0] == "beta": pointer = getattr(pointer, "bias") - elif l[0] == "output_weights": + elif scope_names[0] == "output_weights": pointer = getattr(pointer, "weight") - elif l[0] == "squad": + elif scope_names[0] == "squad": pointer = getattr(pointer, "classifier") else: try: - pointer = getattr(pointer, l[0]) + pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info("Skipping {}".format("/".join(name))) continue - if len(l) >= 2: - num = int(l[1]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] if m_name[-11:] == "_embeddings": pointer = getattr(pointer, "weight") diff --git a/transformers/modeling_albert.py b/transformers/modeling_albert.py index 669d8fc357..5162a1d1de 100644 --- a/transformers/modeling_albert.py +++ b/transformers/modeling_albert.py @@ -124,26 +124,26 @@ def load_tf_weights_in_albert(model, config, tf_checkpoint_path): pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): - l = re.split(r"_(\d+)", m_name) + scope_names = re.split(r"_(\d+)", m_name) else: - l = [m_name] + scope_names = [m_name] - if l[0] == "kernel" or l[0] == "gamma": + if scope_names[0] == "kernel" or scope_names[0] == "gamma": pointer = getattr(pointer, "weight") - elif l[0] == "output_bias" or l[0] == "beta": + elif scope_names[0] == "output_bias" or scope_names[0] == "beta": pointer = getattr(pointer, "bias") - elif l[0] == "output_weights": + elif scope_names[0] == "output_weights": pointer = getattr(pointer, "weight") - elif l[0] == "squad": + elif scope_names[0] == "squad": pointer = getattr(pointer, "classifier") else: try: - pointer = getattr(pointer, l[0]) + pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info("Skipping {}".format("/".join(name))) continue - if len(l) >= 2: - num = int(l[1]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] if m_name[-11:] == "_embeddings": diff --git a/transformers/modeling_bert.py b/transformers/modeling_bert.py index 9c6cccf71f..e2e4f2b5fd 100644 --- a/transformers/modeling_bert.py +++ b/transformers/modeling_bert.py @@ -93,25 +93,25 @@ def load_tf_weights_in_bert(model, config, tf_checkpoint_path): pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): - l = re.split(r"_(\d+)", m_name) + scope_names = re.split(r"_(\d+)", m_name) else: - l = [m_name] - if l[0] == "kernel" or l[0] == "gamma": + scope_names = [m_name] + if scope_names[0] == "kernel" or scope_names[0] == "gamma": pointer = getattr(pointer, "weight") - elif l[0] == "output_bias" or l[0] == "beta": + elif scope_names[0] == "output_bias" or scope_names[0] == "beta": pointer = getattr(pointer, "bias") - elif l[0] == "output_weights": + elif scope_names[0] == "output_weights": pointer = getattr(pointer, "weight") - elif l[0] == "squad": + elif scope_names[0] == "squad": pointer = getattr(pointer, "classifier") else: try: - pointer = getattr(pointer, l[0]) + pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info("Skipping {}".format("/".join(name))) continue - if len(l) >= 2: - num = int(l[1]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] if m_name[-11:] == "_embeddings": pointer = getattr(pointer, "weight") diff --git a/transformers/modeling_gpt2.py b/transformers/modeling_gpt2.py index 1e5c54c95d..ad086c451f 100644 --- a/transformers/modeling_gpt2.py +++ b/transformers/modeling_gpt2.py @@ -77,20 +77,20 @@ def load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path): pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+\d+", m_name): - l = re.split(r"(\d+)", m_name) + scope_names = re.split(r"(\d+)", m_name) else: - l = [m_name] - if l[0] == "w" or l[0] == "g": + scope_names = [m_name] + if scope_names[0] == "w" or scope_names[0] == "g": pointer = getattr(pointer, "weight") - elif l[0] == "b": + elif scope_names[0] == "b": pointer = getattr(pointer, "bias") - elif l[0] == "wpe" or l[0] == "wte": - pointer = getattr(pointer, l[0]) + elif scope_names[0] == "wpe" or scope_names[0] == "wte": + pointer = getattr(pointer, scope_names[0]) pointer = getattr(pointer, "weight") else: - pointer = getattr(pointer, l[0]) - if len(l) >= 2: - num = int(l[1]) + pointer = getattr(pointer, scope_names[0]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] try: assert pointer.shape == array.shape diff --git a/transformers/modeling_openai.py b/transformers/modeling_openai.py index 3f37a4acfb..66487755c6 100644 --- a/transformers/modeling_openai.py +++ b/transformers/modeling_openai.py @@ -90,19 +90,19 @@ def load_tf_weights_in_openai_gpt(model, config, openai_checkpoint_folder_path): pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+\d+", m_name): - l = re.split(r"(\d+)", m_name) + scope_names = re.split(r"(\d+)", m_name) else: - l = [m_name] - if l[0] == "g": + scope_names = [m_name] + if scope_names[0] == "g": pointer = getattr(pointer, "weight") - elif l[0] == "b": + elif scope_names[0] == "b": pointer = getattr(pointer, "bias") - elif l[0] == "w": + elif scope_names[0] == "w": pointer = getattr(pointer, "weight") else: - pointer = getattr(pointer, l[0]) - if len(l) >= 2: - num = int(l[1]) + pointer = getattr(pointer, scope_names[0]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] try: assert pointer.shape == array.shape diff --git a/transformers/modeling_t5.py b/transformers/modeling_t5.py index a1024d47dd..199ec74225 100644 --- a/transformers/modeling_t5.py +++ b/transformers/modeling_t5.py @@ -95,29 +95,29 @@ def load_tf_weights_in_t5(model, config, tf_checkpoint_path): array = tf_weights[txt_name] for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): - l = re.split(r"_(\d+)", m_name) + scope_names = re.split(r"_(\d+)", m_name) else: - l = [m_name] - if l[0] in ["kernel", "scale", "embedding"]: + scope_names = [m_name] + if scope_names[0] in ["kernel", "scale", "embedding"]: pointer = getattr(pointer, "weight") - # elif l[0] == 'scale': + # elif scope_names[0] == 'scale': # pointer = getattr(pointer, 'weight') - # elif l[0] == 'output_bias' or l[0] == 'beta': + # elif scope_names[0] == 'output_bias' or scope_names[0] == 'beta': # pointer = getattr(pointer, 'bias') - # elif l[0] == 'squad': + # elif scope_names[0] == 'squad': # pointer = getattr(pointer, 'classifier') else: try: - pointer = getattr(pointer, l[0]) + pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info("Skipping {}".format("/".join(name))) continue - if len(l) >= 2: - num = int(l[1]) + if len(scope_names) >= 2: + num = int(scope_names[1]) pointer = pointer[num] - if l[0] not in ["kernel", "scale", "embedding"]: + if scope_names[0] not in ["kernel", "scale", "embedding"]: pointer = getattr(pointer, "weight") - if l[0] != "embedding": + if scope_names[0] != "embedding": logger.info("Transposing numpy weight of shape {} for {}".format(array.shape, name)) array = np.transpose(array) try: diff --git a/transformers/tokenization_t5.py b/transformers/tokenization_t5.py index 8eb589cd1c..e9921fef8d 100644 --- a/transformers/tokenization_t5.py +++ b/transformers/tokenization_t5.py @@ -160,8 +160,8 @@ class T5Tokenizer(PreTrainedTokenizer): def _convert_token_to_id(self, token): """ Converts a token (str/unicode) in an id using the vocab. """ if token.startswith("", token) - num = int(l.group(1)) + match = re.match(r"", token) + num = int(match.group(1)) return self.vocab_size - num - 1 return self.sp_model.piece_to_id(token)