Skip to content

Instantly share code, notes, and snippets.

@yngtodd
Created January 15, 2020 16:02
Show Gist options
  • Save yngtodd/f3bda25503a9611765ab33c1178db48c to your computer and use it in GitHub Desktop.
Save yngtodd/f3bda25503a9611765ab33c1178db48c to your computer and use it in GitHub Desktop.
TF Attention
Display the source blob
Display the rendered blob
Raw
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import tensorflow as tf\n",
"from tensorflow.keras.layers import Attention"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [],
"source": [
"attn = tf.keras.layers.Attention(use_scale=True)"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
"x = tf.random.uniform((1, 2, 2))"
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {},
"outputs": [],
"source": [
"def call(q, v, k, mask_q=None, mask_v=None):\n",
" \"\"\" Call attention instance \"\"\"\n",
" attn = tf.keras.layers.Attention(use_scale=True)\n",
" return attn(inputs=[q, v, k], mask=[mask_q, mask_v])"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<tf.Tensor: id=123, shape=(1, 2, 2), dtype=float32, numpy=\n",
"array([[[0.62968266, 0.6612503 ],\n",
" [0.6235384 , 0.73767066]]], dtype=float32)>"
]
},
"execution_count": 27,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"call(x, x, x)"
]
},
{
"cell_type": "code",
"execution_count": 28,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<tf.Tensor: id=135, shape=(1, 2, 2), dtype=float32, numpy=\n",
"array([[[0.62968266, 0.6612503 ],\n",
" [0.6235384 , 0.73767066]]], dtype=float32)>"
]
},
"execution_count": 28,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"call(q=x, v=x, k=x)"
]
},
{
"cell_type": "code",
"execution_count": 29,
"metadata": {},
"outputs": [],
"source": [
"class MyAttention(tf.keras.Model):\n",
" \n",
" def __init__(self):\n",
" super(MyAttention, self).__init__()\n",
" self.attention = Attention(use_scale=True)\n",
" \n",
" def call(self, q, v, k, mask_q=None, mask_v=None):\n",
" return self.attention(inputs=[q, v, k], mask=[mask_q, mask_v])"
]
},
{
"cell_type": "code",
"execution_count": 30,
"metadata": {},
"outputs": [],
"source": [
"needs_attention = NeedsAttention()"
]
},
{
"cell_type": "code",
"execution_count": 31,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<tf.Tensor: id=147, shape=(1, 2, 2), dtype=float32, numpy=\n",
"array([[[0.62968266, 0.6612503 ],\n",
" [0.6235384 , 0.73767066]]], dtype=float32)>"
]
},
"execution_count": 31,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"needs_attention(x, x, x)"
]
},
{
"cell_type": "code",
"execution_count": 33,
"metadata": {},
"outputs": [
{
"ename": "TypeError",
"evalue": "__call__() missing 1 required positional argument: 'inputs'",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mTypeError\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m<ipython-input-33-5fa3b47998d9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mneeds_attention\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mq\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mv\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mk\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
"\u001b[0;31mTypeError\u001b[0m: __call__() missing 1 required positional argument: 'inputs'"
]
}
],
"source": [
"needs_attention(q=x, v=x, k=x)"
]
},
{
"cell_type": "code",
"execution_count": 34,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<tf.Tensor: id=157, shape=(1, 2, 2), dtype=float32, numpy=\n",
"array([[[0.62968266, 0.6612503 ],\n",
" [0.6235384 , 0.73767066]]], dtype=float32)>"
]
},
"execution_count": 34,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"needs_attention.call(q=x, v=x, k=x)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.4"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment