testing_fa_en_models.py

from transformers import MT5Config, MT5ForConditionalGeneration, MT5Tokenizer
from transformers.models.t5.modeling_t5 import load_tf_weights_in_t5

model_name = "persiannlp/mt5-base-parsinlu-opus-translation_fa_en"
tokenizer = MT5Tokenizer.from_pretrained(model_name)
model = MT5ForConditionalGeneration.from_pretrained(model_name)

def run_model(input_string, **generator_args):
    input_ids = tokenizer.encode(input_string, return_tensors="pt")
    res = model.generate(input_ids, **generator_args)
    output = tokenizer.batch_decode(res, skip_special_tokens=True)
    print(output)
    return output

run_model("من در ریاضیات خوب هستم") # your query 
run_model("ستایش خدای را که پروردگار جهانیان است.")
run_model("در هاید پارک کرنر بر گلدانی ایستاده موعظه می‌کند؛")
run_model("وی از تمامی بلاگرها، سازمان‌ها و افرادی که از وی پشتیبانی کرده‌اند، تشکر کرد.")
run_model("مشابه سال ۲۰۰۱، تولید آمونیاک بی آب در ایالات متحده در سال ۲۰۰۰ تقریباً ۱۷،۴۰۰،۰۰۰ تن (معادل بدون آب) با مصرف ظاهری ۲۲،۰۰۰،۰۰۰ تن و حدود ۴۶۰۰۰۰۰ با واردات خالص مواجه شد. ")
run_model("می خواهم دکترای علوم کامپیوتر راجع به شبکه های اجتماعی را دنبال کنم، چالش حل نشده در شبکه های اجتماعی چیست؟")

# this gives me the followings: 
# ['I am well in math']
# ['the admiration of God, which is the Lord of the world.']
# ['At the Ford Park, the Crawford Park stands on a vase;']
# ['He thanked all the bloggers, the organizations, and the people who supported him']
# ['similar to the year 2001, the economy of ammonia in the United States in the']
# ['I want to follow the computer experts on social networks, what is the unsolved problem in']


# if I change the size to "base", I get the following: 
# ["I'm good at math."]
# ['Adoration of God, the Lord of the world.']
# ['At the High End of the Park, Conrad stands on a vase preaching;']
# ['She thanked all the bloggers, organizations, and men who had supported her.']
# ['In 2000, the lack of water ammonia in the United States was almost']
# ['I want to follow the computer science doctorate on social networks. What is the unsolved challenge']


# if I change the size to "large", I get the following: 
# ["I'm good at math"]
# ['the praise of God, the Lord of the world.']
# ['At the Hyde Park Corner, Carpenter is preaching on a vase;']
# ['He thanked all the bloggers, organizations, and people who had supported him.']
# ['Similarly in 2001, the production of waterless ammonia in the United States was']
# ['I want to pursue my degree in Computer Science on social networks, what is the']

Where is the language marker? How does the model know the input and output language?