FM / transformers

/usr/local/lib/python3.6/site-packages/transformers/

/usr/local/lib/python3.6/site-packages/transformers

Name	Size	Mode	Actions
benchmark/	-	0755	rm
commands/	-	0755	rm
data/	-	0755	rm
models/	-	0755	rm
onnx/	-	0755	rm
pipelines/	-	0755	rm
sagemaker/	-	0755	rm
utils/	-	0755	rm
__pycache__/	-	0755	rm
activations.py	6612	0644	edit dl rm
activations_tf.py	4268	0644	edit dl rm
configuration_utils.py	46765	0644	edit dl rm
convert_graph_to_onnx.py	19575	0644	edit dl rm
convert_pytorch_checkpoint_to_tf2.py	16640	0644	edit dl rm
convert_slow_tokenizer.py	37626	0644	edit dl rm
convert_slow_tokenizers_checkpoints_to_fast.py	4954	0644	edit dl rm
convert_tf_hub_seq_to_seq_bert_to_pytorch.py	2899	0644	edit dl rm
debug_utils.py	12887	0644	edit dl rm
deepspeed.py	18883	0644	edit dl rm
dependency_versions_check.py	1767	0644	edit dl rm
dependency_versions_table.py	2390	0644	edit dl rm
dynamic_module_utils.py	19306	0644	edit dl rm
feature_extraction_sequence_utils.py	18074	0644	edit dl rm
feature_extraction_utils.py	26768	0644	edit dl rm
file_utils.py	3930	0644	edit dl rm
generation_beam_constraints.py	19078	0644	edit dl rm
generation_beam_search.py	40145	0644	edit dl rm
generation_flax_logits_process.py	10879	0644	edit dl rm
generation_flax_utils.py	38708	0644	edit dl rm
generation_logits_process.py	30336	0644	edit dl rm
generation_stopping_criteria.py	5509	0644	edit dl rm
generation_tf_logits_process.py	17503	0644	edit dl rm
generation_tf_utils.py	130671	0644	edit dl rm
generation_utils.py	177118	0644	edit dl rm
hf_argparser.py	11901	0644	edit dl rm
image_utils.py	12860	0644	edit dl rm
integrations.py	40203	0644	edit dl rm
keras_callbacks.py	18902	0644	edit dl rm
modelcard.py	35278	0644	edit dl rm
modeling_flax_outputs.py	35514	0644	edit dl rm
modeling_flax_pytorch_utils.py	12296	0644	edit dl rm
modeling_flax_utils.py	37661	0644	edit dl rm
modeling_outputs.py	60327	0644	edit dl rm
modeling_tf_outputs.py	46107	0644	edit dl rm
modeling_tf_pytorch_utils.py	19916	0644	edit dl rm
modeling_tf_utils.py	100832	0644	edit dl rm
modeling_utils.py	137800	0644	edit dl rm
optimization.py	27756	0644	edit dl rm
optimization_tf.py	15722	0644	edit dl rm
processing_utils.py	10559	0644	edit dl rm
py.typed	1	0644	edit dl rm
pytorch_utils.py	1649	0644	edit dl rm
testing_utils.py	48292	0644	edit dl rm
tf_utils.py	1568	0644	edit dl rm
tokenization_utils.py	39867	0644	edit dl rm
tokenization_utils_base.py	170159	0644	edit dl rm
tokenization_utils_fast.py	32909	0644	edit dl rm
trainer.py	143103	0644	edit dl rm
trainer_callback.py	23199	0644	edit dl rm
trainer_pt_utils.py	44764	0644	edit dl rm
trainer_seq2seq.py	10396	0644	edit dl rm
trainer_tf.py	34564	0644	edit dl rm
trainer_utils.py	18509	0644	edit dl rm
training_args.py	68042	0644	edit dl rm
training_args_seq2seq.py	2829	0644	edit dl rm
training_args_tf.py	14267	0644	edit dl rm
__init__.py	171952	0644	edit dl rm

Edit: /usr/local/lib/python3.6/site-packages/transformers/activations_tf.py (4268B)

# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import math

import tensorflow as tf
from packaging import version

def _gelu(x):
    """
    Gaussian Error Linear Unit. Original Implementation of the gelu activation function in Google Bert repo when
    initially created. For information: OpenAI GPT's gelu is slightly different (and gives slightly different results):
    0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3)))) Also see
    https://arxiv.org/abs/1606.08415
    """
    x = tf.convert_to_tensor(x)
    cdf = 0.5 * (1.0 + tf.math.erf(x / tf.cast(tf.sqrt(2.0), x.dtype)))

return x * cdf

def _gelu_new(x):
    """
    Gaussian Error Linear Unit. This is a smoother version of the GELU. Original paper: https://arxiv.org/abs/1606.0841

Args:
        x: float Tensor to perform activation

Returns:
        `x` with the GELU activation applied.
    """
    x = tf.convert_to_tensor(x)
    pi = tf.cast(math.pi, x.dtype)
    coeff = tf.cast(0.044715, x.dtype)
    cdf = 0.5 * (1.0 + tf.tanh(tf.sqrt(2.0 / pi) * (x + coeff * tf.pow(x, 3))))

return x * cdf

def mish(x):
    x = tf.convert_to_tensor(x)

return x * tf.tanh(tf.math.softplus(x))

def gelu_fast(x):
    x = tf.convert_to_tensor(x)
    coeff1 = tf.cast(0.044715, x.dtype)
    coeff2 = tf.cast(0.7978845608, x.dtype)

return 0.5 * x * (1.0 + tf.tanh(x * coeff2 * (1.0 + coeff1 * x * x)))

def quick_gelu(x):
    x = tf.convert_to_tensor(x)
    coeff = tf.cast(1.702, x.dtype)
    return x * tf.math.sigmoid(coeff * x)

def gelu_10(x):
    """
    Clip the range of possible GeLU outputs between [-10, 10]. This is especially useful for quantization purpose, as
    it allows mapping 2 negatives values in the GeLU spectrum. For more information on this trick, please refer to
    https://arxiv.org/abs/2004.09602

Gaussian Error Linear Unit. Original Implementation of the gelu activation function in Google Bert repo when
    initially created. For information: OpenAI GPT's gelu is slightly different (and gives slightly different results):
    0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3)))) Also see
    https://arxiv.org/abs/1606.08415 :param x: :return:
    """
    return tf.clip_by_value(_gelu(x), -10, 10)

def glu(x, axis=-1):
    """
    Gated Linear Unit. Implementation as defined in the original paper (see https://arxiv.org/abs/1612.08083), where
    the input `x` is split in two halves across a dimension (`axis`), A and B, returning A * sigmoid(B).

Args:
        `x`: float Tensor to perform activation
        `axis`: dimension across which `x` be split in half

Returns:
        `x` with the GLU activation applied (with its size halved across the dimension `axis`).
    """
    a, b = tf.split(x, 2, axis=axis)
    return a * tf.math.sigmoid(b)

if version.parse(tf.version.VERSION) >= version.parse("2.4"):

def approximate_gelu_wrap(x):
        return tf.keras.activations.gelu(x, approximate=True)

gelu = tf.keras.activations.gelu
    gelu_new = approximate_gelu_wrap
else:
    gelu = _gelu
    gelu_new = _gelu_new

ACT2FN = {
    "gelu": gelu,
    "relu": tf.keras.activations.relu,
    "swish": tf.keras.activations.swish,
    "silu": tf.keras.activations.swish,
    "gelu_new": gelu_new,
    "mish": mish,
    "tanh": tf.keras.activations.tanh,
    "gelu_fast": gelu_fast,
    "quick_gelu": quick_gelu,
    "gelu_10": gelu_10,
    "glu": glu,
}

def get_tf_activation(activation_string):
    if activation_string in ACT2FN:
        return ACT2FN[activation_string]
    else:
        raise KeyError(f"function {activation_string} not found in ACT2FN mapping {list(ACT2FN.keys())}")

cmd: