EngineX-Cambricon/enginex-mlu370-any2any
9
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init

Browse Source
This commit is contained in:
luopingyi
2025-10-09 16:47:16 +08:00
parent c8feb4deb5
commit e27e3f16bb
5248 changed files with 1778505 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

0
transformers/tests/models/cohere/__init__.py Normal file
View File

250
transformers/tests/models/cohere/test_modeling_cohere.py Normal file
View File

@@ -0,0 +1,250 @@
# Copyright 2022 The HuggingFace Inc. 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.
"""Testing suite for the PyTorch Cohere model."""
import unittest
from transformers import CohereConfig, is_torch_available
from transformers.testing_utils import (
require_bitsandbytes,
require_torch,
require_torch_multi_accelerator,
slow,
torch_device,
)
from ...generation.test_utils import GenerationTesterMixin
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor
from ...test_pipeline_mixin import PipelineTesterMixin
if is_torch_available():
import torch
from transformers import AutoTokenizer, CohereForCausalLM, CohereModel
# Copied from transformers.tests.models.llama.LlamaModelTester with Llama->Cohere
class CohereModelTester:
config_class = CohereConfig
if is_torch_available():
model_class = CohereModel
for_causal_lm_class = CohereForCausalLM
def __init__(
self,
parent,
batch_size=13,
seq_length=7,
is_training=True,
use_input_mask=True,
use_token_type_ids=False,
use_labels=True,
vocab_size=99,
hidden_size=32,
num_hidden_layers=2,
num_attention_heads=4,
intermediate_size=37,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=16,
type_sequence_label_size=2,
initializer_range=0.02,
num_labels=3,
num_choices=4,
pad_token_id=0,
scope=None,
):
self.parent = parent
self.batch_size = batch_size
self.seq_length = seq_length
self.is_training = is_training
self.use_input_mask = use_input_mask
self.use_token_type_ids = use_token_type_ids
self.use_labels = use_labels
self.vocab_size = vocab_size
self.hidden_size = hidden_size
self.num_hidden_layers = num_hidden_layers
self.num_attention_heads = num_attention_heads
self.intermediate_size = intermediate_size
self.hidden_act = hidden_act
self.hidden_dropout_prob = hidden_dropout_prob
self.attention_probs_dropout_prob = attention_probs_dropout_prob
self.max_position_embeddings = max_position_embeddings
self.type_vocab_size = type_vocab_size
self.type_sequence_label_size = type_sequence_label_size
self.initializer_range = initializer_range
self.num_labels = num_labels
self.num_choices = num_choices
self.pad_token_id = pad_token_id
self.scope = scope
def prepare_config_and_inputs(self):
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
input_mask = None
if self.use_input_mask:
input_mask = torch.tril(torch.ones_like(input_ids).to(torch_device))
token_type_ids = None
if self.use_token_type_ids:
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
sequence_labels = None
token_labels = None
choice_labels = None
if self.use_labels:
sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
choice_labels = ids_tensor([self.batch_size], self.num_choices)
config = self.get_config()
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
# Ignore copy
def get_config(self):
return self.config_class(
vocab_size=self.vocab_size,
hidden_size=self.hidden_size,
num_hidden_layers=self.num_hidden_layers,
num_attention_heads=self.num_attention_heads,
intermediate_size=self.intermediate_size,
hidden_act=self.hidden_act,
hidden_dropout_prob=self.hidden_dropout_prob,
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
max_position_embeddings=self.max_position_embeddings,
type_vocab_size=self.type_vocab_size,
is_decoder=False,
initializer_range=self.initializer_range,
pad_token_id=self.pad_token_id,
)
def create_and_check_model(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = self.model_class(config=config)
model.to(torch_device)
model.eval()
result = model(input_ids, attention_mask=input_mask)
result = model(input_ids)
self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
(
config,
input_ids,
token_type_ids,
input_mask,
sequence_labels,
token_labels,
choice_labels,
) = config_and_inputs
inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
return config, inputs_dict
@require_torch
class CohereModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin, unittest.TestCase):
all_model_classes = (CohereModel, CohereForCausalLM) if is_torch_available() else ()
pipeline_model_mapping = (
{
"feature-extraction": CohereModel,
"text-generation": CohereForCausalLM,
}
if is_torch_available()
else {}
)
test_headmasking = False
test_pruning = False
fx_compatible = False
# Need to use `0.8` instead of `0.9` for `test_cpu_offload`
# This is because we are hitting edge cases with the causal_mask buffer
model_split_percents = [0.5, 0.7, 0.8]
def setUp(self):
self.model_tester = CohereModelTester(self)
self.config_tester = ConfigTester(self, config_class=CohereConfig, hidden_size=37)
def test_config(self):
self.config_tester.run_common_tests()
def test_model(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_model(*config_and_inputs)
def test_model_various_embeddings(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
for type in ["absolute", "relative_key", "relative_key_query"]:
config_and_inputs[0].position_embedding_type = type
self.model_tester.create_and_check_model(*config_and_inputs)
def test_torch_fx_output_loss(self):
super().test_torch_fx_output_loss()
@require_torch
@slow
class CohereIntegrationTest(unittest.TestCase):
@require_torch_multi_accelerator
@require_bitsandbytes
def test_batched_4bit(self):
model_id = "CohereForAI/c4ai-command-r-v01-4bit"
EXPECTED_TEXT = [
'Hello today I am going to show you how to make a simple and easy card using the new stamp set called "Hello" from the Occasions catalog. This set is so versatile and can be used for many occasions. I used the new In',
"Hi there, here we are again with another great collection of free fonts for your next project. This time we have gathered 10 free fonts that you can download and use in your designs. These fonts are perfect for any kind",
]
model = CohereForCausalLM.from_pretrained(model_id, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_id)
tokenizer.pad_token = tokenizer.eos_token
text = ["Hello today I am going to show you how to", "Hi there, here we are"]
inputs = tokenizer(text, return_tensors="pt", padding=True).to(torch_device)
output = model.generate(**inputs, max_new_tokens=40, do_sample=False)
self.assertEqual(tokenizer.batch_decode(output, skip_special_tokens=True), EXPECTED_TEXT)
def test_batched_small_model_logits(self):
# Since the model is very large, we created a random cohere model so that we can do a simple
# logits check on it.
model_id = "hf-internal-testing/cohere-random"
EXPECTED_LOGITS = torch.Tensor(
[
[[0.0000, 0.0285, 0.0322], [0.0000, 0.0011, 0.1105], [0.0000, -0.0018, -0.1019]],
[[0.0000, 0.1080, 0.0454], [0.0000, -0.1808, -0.1553], [0.0000, 0.0452, 0.0369]],
]
).to(device=torch_device, dtype=torch.float16)
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = CohereForCausalLM.from_pretrained(model_id, dtype=torch.float16).to(torch_device)
tokenizer.pad_token = tokenizer.eos_token
text = ["Hello today I am going to show you how to", "Hi there, here we are"]
inputs = tokenizer(text, return_tensors="pt", padding=True).to(torch_device)
with torch.no_grad():
output = model(**inputs)
logits = output.logits
torch.testing.assert_close(EXPECTED_LOGITS, logits[:, -3:, :3], rtol=1e-3, atol=1e-3)

297
transformers/tests/models/cohere/test_tokenization_cohere.py Normal file
View File

@@ -0,0 +1,297 @@
# Copyright 2022 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 copy
import unittest
from transformers import CohereTokenizerFast
from transformers.testing_utils import (
require_jinja,
require_tokenizers,
require_torch_multi_accelerator,
)
from ...test_tokenization_common import TokenizerTesterMixin
@require_tokenizers
class CohereTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
slow_tokenizer_class = None
rust_tokenizer_class = CohereTokenizerFast
tokenizer_class = CohereTokenizerFast
test_rust_tokenizer = True
test_slow_tokenizer = False
from_pretrained_vocab_key = "tokenizer_file"
from_pretrained_id = "hf-internal-testing/tiny-random-CohereForCausalLM"
special_tokens_map = {
"bos_token": "<BOS_TOKEN>",
"eos_token": "<|END_OF_TURN_TOKEN|>",
"unk_token": "<UNK>",
"pad_token": "<PAD>",
}
@classmethod
def setUpClass(cls):
super().setUpClass()
tokenizer = CohereTokenizerFast.from_pretrained("hf-internal-testing/tiny-random-CohereForCausalLM")
tokenizer.save_pretrained(cls.tmpdirname)
@classmethod
def get_rust_tokenizer(cls, pretrained_name=None, **kwargs):
_kwargs = copy.deepcopy(cls.special_tokens_map)
_kwargs.update(kwargs)
kwargs = _kwargs
pretrained_name = pretrained_name or cls.tmpdirname
return CohereTokenizerFast.from_pretrained(pretrained_name, **kwargs)
# This gives CPU OOM on a single-gpu runner (~60G RAM). On multi-gpu runner, it has ~180G RAM which is enough.
@require_torch_multi_accelerator
def test_torch_encode_plus_sent_to_model(self):
super().test_torch_encode_plus_sent_to_model()
@unittest.skip(reason="This needs a slow tokenizer. Cohere does not have one!")
def test_encode_decode_with_spaces(self):
return
def test_encodings_from_sample_data(self):
"""
Assert that the created tokens are the same than the hard-coded ones
"""
tokenizer = self.get_rust_tokenizer()
INPUT_SENTENCES = ["The quick brown fox<|END_OF_TURN_TOKEN|>", "jumps over the lazy dog<|END_OF_TURN_TOKEN|>"]
TARGET_TOKENS = [
[5, 60, 203, 746, 666, 980, 571, 222, 87, 96, 8],
[5, 82, 332, 88, 91, 544, 206, 257, 930, 97, 239, 435, 8],
]
computed_tokens = tokenizer.batch_encode_plus(INPUT_SENTENCES)["input_ids"]
self.assertListEqual(TARGET_TOKENS, computed_tokens)
INPUT_SENTENCES_W_BOS = [
"<BOS_TOKEN>The quick brown fox<|END_OF_TURN_TOKEN|>",
"<BOS_TOKEN>jumps over the lazy dog<|END_OF_TURN_TOKEN|>",
]
decoded_tokens = tokenizer.batch_decode(computed_tokens)
self.assertListEqual(decoded_tokens, INPUT_SENTENCES_W_BOS)
def test_padding(self, max_length=10):
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
with self.subTest(f"{tokenizer.__class__.__name__} ({pretrained_name})"):
tokenizer_r = self.get_rust_tokenizer(pretrained_name, **kwargs)
# tokenizer_r.pad_token = None # Hotfixing padding = None
# Simple input
s = "This is a simple input"
s2 = ["This is a simple input 1", "This is a simple input 2"]
p = ("This is a simple input", "This is a pair")
p2 = [
("This is a simple input 1", "This is a simple input 2"),
("This is a simple pair 1", "This is a simple pair 2"),
]
# Simple input tests
try:
tokenizer_r.encode(s, max_length=max_length)
tokenizer_r.encode_plus(s, max_length=max_length)
tokenizer_r.batch_encode_plus(s2, max_length=max_length)
tokenizer_r.encode(p, max_length=max_length)
tokenizer_r.batch_encode_plus(p2, max_length=max_length)
except ValueError:
self.fail("Cohere Tokenizer should be able to deal with padding")
tokenizer_r.pad_token = None # Hotfixing padding = None
self.assertRaises(ValueError, tokenizer_r.encode, s, max_length=max_length, padding="max_length")
# Simple input
self.assertRaises(ValueError, tokenizer_r.encode_plus, s, max_length=max_length, padding="max_length")
# Simple input
self.assertRaises(
ValueError,
tokenizer_r.batch_encode_plus,
s2,
max_length=max_length,
padding="max_length",
)
# Pair input
self.assertRaises(ValueError, tokenizer_r.encode, p, max_length=max_length, padding="max_length")
# Pair input
self.assertRaises(ValueError, tokenizer_r.encode_plus, p, max_length=max_length, padding="max_length")
# Pair input
self.assertRaises(
ValueError,
tokenizer_r.batch_encode_plus,
p2,
max_length=max_length,
padding="max_length",
)
def test_pretrained_model_lists(self):
# No `max_model_input_sizes` for Cohere model
self.assertGreaterEqual(len(self.tokenizer_class.pretrained_vocab_files_map), 1)
self.assertGreaterEqual(len(list(self.tokenizer_class.pretrained_vocab_files_map.values())[0]), 1)
@require_jinja
def test_tokenization_for_chat(self):
tokenizer = self.get_rust_tokenizer()
test_chats = [
[{"role": "system", "content": "You are a helpful chatbot."}, {"role": "user", "content": "Hello!"}],
[
{"role": "system", "content": "You are a helpful chatbot."},
{"role": "user", "content": "Hello!"},
{"role": "assistant", "content": "Nice to meet you."},
],
]
tokenized_chats = [tokenizer.apply_chat_template(test_chat) for test_chat in test_chats]
# fmt: off
expected_tokens = [
[5, 36, 99, 59, 60, 41, 58, 60, 71, 55, 46, 71, 60, 61, 58, 54, 71, 60, 55, 51, 45, 54, 99, 38, 36, 99, 59, 65, 59, 60, 45, 53, 71, 60, 55, 51, 45, 54, 99, 38, 65, 243, 394, 204, 336, 84, 88, 887, 374, 216, 74, 286, 22, 8, 36, 99, 59, 60, 41, 58, 60, 71, 55, 46, 71, 60, 61, 58, 54, 71, 60, 55, 51, 45, 54, 99, 38, 36, 99, 61, 59, 45, 58, 71, 60, 55, 51, 45, 54, 99, 38, 48, 420, 87, 9, 8],
[5, 36, 99, 59, 60, 41, 58, 60, 71, 55, 46, 71, 60, 61, 58, 54, 71, 60, 55, 51, 45, 54, 99, 38, 36, 99, 59, 65,
59, 60, 45, 53, 71, 60, 55, 51, 45, 54, 99, 38, 65, 243, 394, 204, 336, 84, 88, 887, 374, 216, 74, 286, 22, 8,
36, 99, 59, 60, 41, 58, 60, 71, 55, 46, 71, 60, 61, 58, 54, 71, 60, 55, 51, 45, 54, 99, 38, 36, 99, 61, 59,
45, 58, 71, 60, 55, 51, 45, 54, 99, 38, 48, 420, 87, 9, 8, 36, 99, 59, 60, 41, 58, 60, 71, 55, 46, 71, 60, 61,
58, 54, 71, 60, 55, 51, 45, 54, 99, 38, 36, 99, 43, 48, 41, 60, 42, 55, 60, 71, 60, 55, 51, 45, 54, 99, 38,
54, 567, 235, 693, 276, 411, 243, 22, 8]
]
# fmt: on
for tokenized_chat, expected_tokens in zip(tokenized_chats, expected_tokens):
self.assertListEqual(tokenized_chat, expected_tokens)
@require_jinja
def test_tokenization_for_tool_use(self):
tokenizer = self.get_rust_tokenizer()
conversation = [{"role": "user", "content": "Whats the biggest penguin in the world?"}]
tools = [
{
"name": "internet_search",
"description": "Returns a list of relevant document snippets for a textual query retrieved from the internet",
"parameter_definitions": {
"query": {"description": "Query to search the internet with", "type": "str", "required": True}
},
},
{
"name": "directly_answer",
"description": "Calls a standard (un-augmented) AI chatbot to generate a response given the conversation history",
"parameter_definitions": {},
},
]
tool_use_prompt = tokenizer.apply_tool_use_template(
conversation,
tools=tools,
tokenize=False,
add_generation_prompt=True,
)
expected_prompt = '''<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|># Safety Preamble
The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
# System Preamble
## Basic Rules
You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
# User Preamble
## Task and Context
You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.
## Style Guide
Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.
## Available Tools
Here is a list of tools that you have available to you:
```python
def internet_search(query: str) -> List[Dict]:
"""Returns a list of relevant document snippets for a textual query retrieved from the internet
Args:
query (str): Query to search the internet with
"""
pass
```
```python
def directly_answer() -> List[Dict]:
"""Calls a standard (un-augmented) AI chatbot to generate a response given the conversation history
"""
pass
```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Whats the biggest penguin in the world?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Write 'Action:' followed by a json-formatted list of actions that you want to perform in order to produce a good response to the user's last input. You can use any of the supplied tools any number of times, but you should aim to execute the minimum number of necessary actions for the input. You should use the `directly-answer` tool if calling the other tools is unnecessary. The list of actions you want to call should be formatted as a list of json objects, for example:
```json
[
{
"tool_name": title of the tool in the specification,
"parameters": a dict of parameters to input into the tool as they are defined in the specs, or {} if it takes no parameters
}
]```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>'''
self.assertEqual(tool_use_prompt, expected_prompt)
@require_jinja
def test_tokenization_for_grounded_generation(self):
tokenizer = self.get_rust_tokenizer()
conversation = [{"role": "user", "content": "Whats the biggest penguin in the world?"}]
documents = [
{"title": "Tall penguins", "text": "Emperor penguins are the tallest growing up to 122 cm in height."},
{"title": "Penguin habitats", "text": "Emperor penguins only live in Antarctica."},
]
grounded_generation_prompt = tokenizer.apply_grounded_generation_template(
conversation,
documents=documents,
citation_mode="accurate", # or "fast"
tokenize=False,
add_generation_prompt=True,
)
expected_prompt = """<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|># Safety Preamble
The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
# System Preamble
## Basic Rules
You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
# User Preamble
## Task and Context
You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.
## Style Guide
Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Whats the biggest penguin in the world?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><results>
Document: 0
title: Tall penguins
text: Emperor penguins are the tallest growing up to 122 cm in height.
Document: 1
title: Penguin habitats
text: Emperor penguins only live in Antarctica.
</results><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Carefully perform the following instructions, in order, starting each with a new line.
Firstly, Decide which of the retrieved documents are relevant to the user's last input by writing 'Relevant Documents:' followed by comma-separated list of document numbers. If none are relevant, you should instead write 'None'.
Secondly, Decide which of the retrieved documents contain facts that should be cited in a good answer to the user's last input by writing 'Cited Documents:' followed a comma-separated list of document numbers. If you dont want to cite any of them, you should instead write 'None'.
Thirdly, Write 'Answer:' followed by a response to the user's last input in high quality natural english. Use the retrieved documents to help you. Do not insert any citations or grounding markup.
Finally, Write 'Grounded answer:' followed by a response to the user's last input in high quality natural english. Use the symbols <co: doc> and </co: doc> to indicate when a fact comes from a document in the search result, e.g <co: 0>my fact</co: 0> for a fact from document 0.<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>"""
self.assertEqual(grounded_generation_prompt, expected_prompt)
def test_add_prefix_space_fast(self):
tokenizer_w_prefix = self.get_rust_tokenizer(add_prefix_space=True)
tokenizer_wo_prefix = self.get_rust_tokenizer(add_prefix_space=False)
tokens_w_prefix = tokenizer_w_prefix.tokenize("Hey")
tokens_wo_prefix = tokenizer_wo_prefix.tokenize("Hey")
self.assertNotEqual(tokens_w_prefix, tokens_wo_prefix)