persiannlp / parsinlu Goto Github PK

The lines are not proper JSON. Unlike Python syntax, JSON strings only use double-quotations: https://stackoverflow.com/a/4162651/1164246

I ran train_and_evaluate_entailment_baselines.sh and got results in /scripts/runs/ . All events.out.tfevents files have same 40 bytes size and are quite empty. How can I solve it?

What is Parsnlu's best model for entailment.

Have you trained 'persiannlp/parsbert-base-parsinlu-entailment' on FarsTail dataset?

Hello,
I am trying to use reading comprehension data in another model. I found out that there are several answers to most of the questions. I tried to concatenate these answers but it seems that some of them complete each other but others become duplicated.
For example:

1. 20 دندان / 10 عدد در هر آرواره
2. استان فارس / فارس

I would like to know how should I use the answers and solve this problem.
Thanks for your consideration.

Hi, first I want to send gratitude for sharing your great work. I'm new to NLP field and I have to a task to do, so I've looked at your model in hugging face and I want to fine-tune your model. So how can I fine-tune your model with python transformer package?
Thanks a lot

I cloned the package and installed requirements.txt

Then I wento the scripts directory and run this command:

./train_and_evaluate_reading_comprehension_baselines.sh

But I got the following error:

Traceback (most recent call last):
  File "../src/run_seq2seq.py", line 344, in <module>
    main(args)
  File "../src/run_seq2seq.py", line 322, in main
    logger=logger,
  File "/home/pouramini/parsinlu/src/lightning_base.py", line 329, in generic_train
    **train_params,
  File "/home/pouramini/miniconda3/lib/python3.7/site-packages/pytorch_lightning/trainer/connectors/env_vars_connector.py", line 41, in overwrite_by_env_vars
    return fn(self, **kwargs)
TypeError: __init__() got an unexpected keyword argument 'early_stop_callback'

It might be related to some enviornment setting or version mismatche, if you know the error please guide me.

Hi,

Does anyone know how to use ParsiNLU with RASA?
https://rasa.com/docs/rasa/language-support/

Regards,

I tried this example for wikibert-base-parsinlu-multiple-choice.

I got these outputs:

MultipleChoiceModelOutput(loss=None, logits=tensor([[2.1362],
        [2.1362],
        [2.1362],
        [2.1362]], grad_fn=<ViewBackward>), hidden_states=None, attentions=None)
MultipleChoiceModelOutput(loss=None, logits=tensor([[2.1362],
        [2.1362],
        [2.1362],
        [2.1362]], grad_fn=<ViewBackward>), hidden_states=None, attentions=None)
MultipleChoiceModelOutput(loss=None, logits=tensor([[2.1362],
        [2.1362],
        [2.1362],
        [2.1362]], grad_fn=<ViewBackward>), hidden_states=None, attentions=None)

But how do I obtain which candidate was guessed? Those outputs did not make a sense.

The ones that involves BERT had same examples except model name so I guess they would output something similar. Only MT5 ones seemed to output what I expected.

Hi,
Thank you for providing this benchmark,

However, I can't find data for reading comprehension in the data folder.

I'm translating English sentences into Farsi with mt5-base-parsinlu-translation_en_fa (from Huggingface). Sentences longer than around 8 words result in the translation of the first part of the sentence, but the rest of the sentence is ignored. For example:

English sentences:

Terry's side fell to their second Premier League loss of the season at Loftus Road

Following a four-day hiatus, UN envoy Ismail Ould Cheikh Ahmed on Thursday will resume mediation efforts in the second round of Kuwait-hosted peace talks between Yemen’s warring rivals.

Mark Woods is a writer and broadcaster who has covered the NBA, and British basketball, for over a decade.

Translations:

طرفدار تری در فوتبال دوم فصل در لئوپوس رود به

پس از چهار روز توقف، سفیر سازمان ملل، ایمیل اولد شیخ

مارک ولز نویسنده و پخش کننده ای است که بیش از یک دهه

which according to Google Translate translates back to this:

More fans in the second football season in Leopard

After a four-day hiatus, the ambassador to the United Nations, Old Sheikh Sheikh

Mark Wells has been a writer and broadcaster for over a decade

I can't find any configuration settings that would be limiting the number of tokens being translated
Here is my code:

#!/usr/bin/python3
import sys
#from transformers import MarianTokenizer, MarianMTModel
from transformers import MT5ForConditionalGeneration, MT5Tokenizer
from typing import List
import torch

device = "cuda:0"

dir=sys.argv[1] + "persiannlp"
size="base"
mname = f'{dir}/data/mt5-{size}-parsinlu-translation_en_fa'

tokenizer = MT5Tokenizer.from_pretrained(mname)
model = MT5ForConditionalGeneration.from_pretrained(mname)
model = model.to(device)

lines = [] 
while True:
    for line in sys.stdin:
        line = line.strip()
        if line == 'EOD':
            inputs    = tokenizer(lines, return_tensors="pt", padding=True).to(device)
            translated   = model.generate(**inputs).to(device)
            [print(tokenizer.decode(t, skip_special_tokens=True)) for t in translated]
            print('EOL')
            sys.stdout.flush()
            lines.clear()
        elif line.startswith('EOF'):
            sys.exit(0)
        else:
            lines.append(line)
sys.exit(0)

I was trying to experiment with your reading comprehension mt5 model and ran into a problem. I followed the code provided in the model card of huggingface
mt5-small-parsinlu.

when I run

tokenizer = MT5Tokenizer.from_pretrained("persiannlp/mt5-small-parsinlu-squad-reading-comprehension")

I get no error but I get the error "AttributeError: 'NoneType' object has no attribute 'encode'" when I try to call the run_model() which points to the tokenizer. I checked the type of tokenizer with both "base" and "small" models and it returns a NoneType. The model, however, loads fine and the type is correct using the same checkpoint name.