paddlepaddle--paddlenlp/tests/transformers/clip/test_tokenizer.py

# coding=utf-8
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
# Copyright 2021 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 json
import os
import unittest

from paddlenlp.transformers import CLIPTokenizer

from ...transformers.test_tokenizer_common import TokenizerTesterMixin

VOCAB_FILES_NAMES = CLIPTokenizer.resource_files_names


class CLIPTokenizationTest(TokenizerTesterMixin, unittest.TestCase):

    tokenizer_class = CLIPTokenizer
    test_rust_tokenizer = True
    from_pretrained_kwargs = {}
    test_seq2seq = False

    def setUp(self):
        super().setUp()

        # fmt: off
        vocab = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n", "lo", "l</w>", "w</w>", "r</w>", "t</w>", "low</w>", "er</w>", "lowest</w>", "newer</w>", "wider", "<unk>", "<|startoftext|>", "<|endoftext|>"]
        # fmt: on
        vocab_tokens = dict(zip(vocab, range(len(vocab))))
        merges = ["#version: 0.2", "l o", "lo w</w>", "e r</w>"]
        self.special_tokens_map = {"unk_token": "<unk>"}

        self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"])
        self.merges_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["merges_file"])
        with open(self.vocab_file, "w", encoding="utf-8") as fp:
            fp.write(json.dumps(vocab_tokens) + "\n")
        with open(self.merges_file, "w", encoding="utf-8") as fp:
            fp.write("\n".join(merges))

    def get_tokenizer(self, **kwargs):
        kwargs.update(self.special_tokens_map)
        if "model_max_length" not in kwargs:
            kwargs.update({"model_max_length": 512})
        return CLIPTokenizer.from_pretrained(self.tmpdirname, **kwargs)

    def get_input_output_texts(self, tokenizer):
        input_text = "lower newer"
        output_text = "lower newer"
        return input_text, output_text

    def test_full_tokenizer(self):
        tokenizer = self.get_tokenizer()
        text = "lower newer"
        bpe_tokens = ["lo", "w", "er</w>", "n", "e", "w", "er</w>"]
        tokens = tokenizer.tokenize(text)
        self.assertListEqual(tokens, bpe_tokens)

        input_tokens = tokens + [tokenizer.unk_token]
        input_bpe_tokens = [10, 2, 16, 9, 3, 2, 16, 20]
        self.assertListEqual(tokenizer.convert_tokens_to_ids(input_tokens), input_bpe_tokens)

    def test_padding_if_pad_token_set_slow(self):
        tokenizer = CLIPTokenizer.from_pretrained(self.tmpdirname, pad_token="<pad>")

        # Simple input
        s = "This is a simple input"
        s2 = ["This is a simple input looooooooong", "This is a simple input"]
        p = ("This is a simple input", "This is a pair")
        p2 = [
            ("This is a simple input loooooong", "This is a simple input"),
            ("This is a simple pair loooooong", "This is a simple pair"),
        ]

        pad_token_id = tokenizer.pad_token_id

        out_s = tokenizer(s, padding="max_length", max_length=30, return_tensors="np", return_attention_mask=True)
        out_s2 = tokenizer(s2, padding=True, truncate=True, return_tensors="np", return_attention_mask=True)
        out_p = tokenizer(*p, padding="max_length", max_length=60, return_tensors="np", return_attention_mask=True)
        out_p2 = tokenizer(p2, padding=True, truncate=True, return_tensors="np", return_attention_mask=True)

        # s
        # test single string max_length padding
        self.assertEqual(out_s["input_ids"].shape[-1], 30)
        self.assertTrue(pad_token_id in out_s["input_ids"])
        self.assertTrue(0 in out_s["attention_mask"])

        # s2
        # test automatic padding
        self.assertEqual(out_s2["input_ids"].shape[-1], 31)
        # long slice doesn't have padding
        self.assertFalse(pad_token_id in out_s2["input_ids"][0])
        self.assertFalse(0 in out_s2["attention_mask"][0])
        # short slice does have padding
        self.assertTrue(pad_token_id in out_s2["input_ids"][1])
        self.assertTrue(0 in out_s2["attention_mask"][1])

        # p
        # test single pair max_length padding
        self.assertEqual(out_p["input_ids"].shape[-1], 60)
        self.assertTrue(pad_token_id in out_p["input_ids"])
        self.assertTrue(0 in out_p["attention_mask"])

        # p2
        # test automatic padding pair
        self.assertEqual(out_p2["input_ids"].shape[-1], 48)
        # long slice pair doesn't have padding
        self.assertFalse(pad_token_id in out_p2["input_ids"][0])
        self.assertFalse(0 in out_p2["attention_mask"][0])
        # short slice pair does have padding
        self.assertTrue(pad_token_id in out_p2["input_ids"][1])
        self.assertTrue(0 in out_p2["attention_mask"][1])

    def test_add_bos_token_slow(self):
        bos_token = "$$$"
        tokenizer = CLIPTokenizer.from_pretrained(self.tmpdirname, bos_token=bos_token, add_bos_token=True)

        s = "This is a simple input"
        s2 = ["This is a simple input 1", "This is a simple input 2"]

        bos_token_id = tokenizer.bos_token_id

        out_s = tokenizer(s)
        out_s2 = tokenizer(s2)

        self.assertEqual(out_s.input_ids[0], bos_token_id)
        self.assertTrue(all(o[0] == bos_token_id for o in out_s2.input_ids))

        decode_s = tokenizer.decode(out_s.input_ids)
        decode_s2 = tokenizer.batch_decode(out_s2.input_ids)

        self.assertEqual(decode_s.split()[0], bos_token)
        self.assertTrue(all(d.split()[0] == bos_token for d in decode_s2))

    @unittest.skip(reason="CLIP always lower cases letters")
    def test_added_tokens_do_lower_case(self):
        # CLIP always lower cases letters
        pass

    @unittest.skip(reason="CLIP do not check pretrained_model_lists")
    def test_pretrained_model_lists(self):
        pass