Quick Fix: The "transformers" library encourages the use of config files. In this case, we need to pass a GenerationConfig object early, rather than to set attributes.
from transformers import AutoTokenizer, BartForConditionalGeneration
model = BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
tokenizer = AutoTokenizer.from_pretrained("facebook/bart-large-cnn")
ARTICLE_TO_SUMMARIZE = (
"PG&E stated it scheduled the blackouts in response to forecasts for high winds "
"amid dry conditions. The aim is to reduce the risk of wildfires. Nearly 800 thousand customers were "
"scheduled to be affected by the shutoffs which were expected to last through at least midday tomorrow."
)
inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors="pt")
# change config and generate summary
from transformers.generation import GenerationConfig
model.config.max_new_tokens = 10
model.config.min_length = 1
gen_cfg = GenerationConfig.from_model_config(model.config)
gen_cfg.max_new_tokens = 10
gen_cfg.min_length = 1
summary_ids = model.generate(inputs["input_ids"], generation_config=gen_cfg)
tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
The Problem:
When training a sequence-to-sequence model using HuggingFace Transformers’ Seq2SeqTrainer, the user encounters a deprecation warning about using a deprecated strategy to control generation. The user wants to modify their code to use the recommended approach but is unable to access the documentation link provided in the warning message. Additionally, the error occurs despite using the latest versions of Transformers (4.28.1) and Python (3.9.7).
The Solutions:
Solution 1: Use Generation Configuration File
The use of configuration files is recommended for controlling generation parameters. To resolve the deprecation warning, create a GenerationConfig object and pass it to the generate method instead of modifying the model’s configuration directly.
In HuggingFace, how can I fix the ‘deprecated strategy’ warning when modifying the model configuration for text generation?
—
Use a ‘GenerationConfig’ object instead of setting configuration attributes directly.
Video Explanation:
The following video, titled "Accelerate Transformer Model Training with Hugging Face and ...", provides additional insights and in-depth exploration related to the topics discussed in this post.
Transformer models deliver state-of-the-art performance on a wide range of machine learning tasks, such as natural language processing, ...
