test of ModernBERT2Olmo-large_1b

experimental seq2seq with EncoderDecoderModel. You will need to patch modeling_llama.py with this code for it work

WIP + output of this model is gibberish bc cross attn needs training

uses different configuration token ids than the first one + uses olmo-1-b-0724 for decoder

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

model_name ="pszemraj/ModernBERT2Olmo-large_1b-cfg2"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

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."
)
prompt = f"summarize dis botmon: {ARTICLE_TO_SUMMARIZE}"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

# autoregressively generate summary (uses greedy decoding by default)
generated_ids = model.generate(
    **inputs,
    min_new_tokens=10,
    max_new_tokens=100,
)
generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(generated_text)