Encoder Decoder
源码 examples/offline_inference/encoder_decoder.py
# SPDX-License-Identifier: Apache-2.0
'''
Demonstrate prompting of text-to-text
encoder/decoder models, specifically BART
'''
from vllm import LLM, SamplingParams
from vllm.inputs import (ExplicitEncoderDecoderPrompt, TextPrompt,
TokensPrompt, zip_enc_dec_prompts)
dtype = "float"
# Create a BART encoder/decoder model instance
# 创建一个 BART 编码器/解码器模型实例
llm = LLM(
model="facebook/bart-large-cnn",
dtype=dtype,
)
# Get BART tokenizer
# 获取 BART tokenizer
tokenizer = llm.llm_engine.get_tokenizer_group()
# Test prompts
#
# This section shows all of the valid ways to prompt an
# encoder/decoder model.
#
# - Helpers for building prompts
# 测试提示
#
# 本节显示了提示的所有有效方法
# 编码器/解码器模型。
#
# - 构建提示的帮助
text_prompt_raw = "Hello, my name is"
text_prompt = TextPrompt(prompt="The president of the United States is")
tokens_prompt = TokensPrompt(prompt_token_ids=tokenizer.encode(
prompt="The capital of France is"))
# - Pass a single prompt to encoder/decoder model
# (implicitly encoder input prompt);
# decoder input prompt is assumed to be None
# - 将单个提示传递到编码器/解码器模型
# (隐式编码器输入提示) ;
# 假定解码器输入提示符为 None
single_text_prompt_raw = text_prompt_raw # Pass a string directly
single_text_prompt = text_prompt # Pass a TextPrompt
single_tokens_prompt = tokens_prompt # Pass a TokensPrompt
# - Pass explicit encoder and decoder input prompts within one data structure.
# Encoder and decoder prompts can both independently be text or tokens, with
# no requirement that they be the same prompt type. Some example prompt-type
# combinations are shown below, note that these are not exhaustive.
# - 在一个数据结构中传递显式编码器和解码器输入提示。
# 编码器和解码器提示可以独立地是文本或 token ，
# 不需要它们是相同的提示类型。一些示例及时类型
# 组合如下所示，请注意，这些并不详尽。
enc_dec_prompt1 = ExplicitEncoderDecoderPrompt(
# Pass encoder prompt string directly, &
# pass decoder prompt tokens
# 直接传递编码器提示字符串，
# 并传递解码器提示 token
encoder_prompt=single_text_prompt_raw,
decoder_prompt=single_tokens_prompt,
)
enc_dec_prompt2 = ExplicitEncoderDecoderPrompt(
# Pass TextPrompt to encoder, and
# pass decoder prompt string directly
# 将 TextPrompt 传递给 encoder
# 直接传递 decoder prompt 字符串
encoder_prompt=single_text_prompt,
decoder_prompt=single_text_prompt_raw,
)
enc_dec_prompt3 = ExplicitEncoderDecoderPrompt(
# Pass encoder prompt tokens directly, and
# pass TextPrompt to decoder
# 通过直接传递编码器提示 token ，然后
# 传递 TextPrompt 到解码器
encoder_prompt=single_tokens_prompt,
decoder_prompt=single_text_prompt,
)
# - Finally, here's a useful helper function for zipping encoder and
# decoder prompts together into a list of ExplicitEncoderDecoderPrompt
# instances
# - 最后，这是用于 zipping 编码器的有用的助手功能，
# 解码器提示一起进入 explicitencoderdecoderprompt 的列表实例
zipped_prompt_list = zip_enc_dec_prompts(
['An encoder prompt', 'Another encoder prompt'],
['A decoder prompt', 'Another decoder prompt'])
# - Let's put all of the above example prompts together into one list
# which we will pass to the encoder/decoder LLM.
# - 让我们将上述所有示例提示放在一个列表中
# 我们将传递给编码器/解码器 LLM。
prompts = [
single_text_prompt_raw, single_text_prompt, single_tokens_prompt,
enc_dec_prompt1, enc_dec_prompt2, enc_dec_prompt3
] + zipped_prompt_list
print(prompts)
# Create a sampling params object.
# 创建一个采样参数对象。
sampling_params = SamplingParams(
temperature=0,
top_p=1.0,
min_tokens=0,
max_tokens=20,
)
# Generate output tokens from the prompts. The output is a list of
# RequestOutput objects that contain the prompt, generated
# text, and other information.
# 从提示中生成输出 token 。
# 输出是包含提示的对象，生成了文本和其他信息。
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
# 打印输出。
for output in outputs:
prompt = output.prompt
encoder_prompt = output.encoder_prompt
generated_text = output.outputs[0].text
print(f"Encoder prompt: {encoder_prompt!r}, "
f"Decoder prompt: {prompt!r}, "
f"Generated text: {generated_text!r}")