README.md

---
datasets:
- IlyaGusev/habr
- Den4ikAI/russian_instructions
- wiki_qa
inference:
  parameters:
    min_length: 20
    max_new_tokens: 250
    top_k: 50
    top_p: 0.9 
    early_stopping: True
    no_repeat_ngram_size: 2
    use_cache: True
    repetition_penalty: 1.5 
    length_penalty: 0.8
    num_beams: 2
        
license: apache-2.0
language:
- ru
pipeline_tag: text-generation
widget:
- text: "Может ли встретиться пингвин и белый медведь?"
  example_title: Question Answering
- text: "Как зарабатывать много денег обучая модели?"
  example_title: Open domain Knoweledge
- text: "Напиши код который выведет Привет Мир"
  example_title: Scientific knowledge
library_name: transformers
tags:
- finance
- code
---

<h1 style="font-size: 42px">Instructions ruGPT Medium v0.11_75к_a<h1/>



# Model Summary

> Я дообучил small rugpt на датасете инструкций, хабра, QA и кода


# Quick Start

```python
from transformers import pipeline
pipe = pipeline(model='AlexWortega/instruct_rugptMedium')
pipe('''Как собрать питон код?''')
```
or
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("AlexWortega/instruct_rugptMedium")
model = AutoModelForCausalLM.from_pretrained("AlexWortega/instruct_rugptMedium")
```
обратите внимание, что лучшие параметры для генерации 
```
gen_kwargs = {
        "min_length": 20,
        "max_new_tokens": 100,
        "top_k": 50,
        "top_p": 0.9,
        "do_sample": True,  
        "early_stopping": True,
        "no_repeat_ngram_size": 2,
        "eos_token_id": tokenizer.eos_token_id,
        "pad_token_id": tokenizer.eos_token_id,
        "use_cache": True,
        "repetition_penalty": 1.5,  
        "length_penalty": 0.8,  
        "num_beams": 4,
        "num_return_sequences": k
    }
```  
# License

The weights of Instructions ruGPT Small v0.1a are licensed under version 2.0 of the Apache License.



## Hyperparameters

I used Novograd with a learning rate of 2e-5 and global batch size of 6 (3 for each data parallel worker).
I use both data parallelism and pipeline parallelism to conduct training.
During training, we truncate the input sequence to 1024 tokens, and for input sequence that contains less than 1024 tokens, we concatenate multiple sequences into one long sequence to improve the data efficiency.



# References

#Metrics

SOON

## BibTeX entry and citation info

```bibtex
@article{
  title={GPT2xl is underrated task solver},
  author={Nickolich Aleksandr, 5Q, datascience, Ilya Gusev, Alex Kukushkin, Karina Romanova, Arseniy Shahmatov, Maksim Gersimenko},
  year={2023}
}
```