alexvumnov/yaml_completion_8bit

Model Card for Model ID

Model finetuned to autocomplete for YAML files 8 bit quantized version of https://huggingface.co/alexvumnov/yaml_completion

Model Details

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

• Developed by: https://huggingface.co/alexvumnov
• Model type: Autoregressive decoder only transformer (GPT2-based)
• Language(s) (NLP): English, but mostly YAML
• License: MIT
• Finetuned from model [optional]: https://huggingface.co/Salesforce/codegen-350M-multi/tree/main

Uses

Model expects a specific prompt format, so please use it for best performance

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

model = AutoModelForCausalLM.from_pretrained("alexvumnov/yaml_completion")
tokenizer = AutoTokenizer.from_pretrained("alexvumnov/yaml_completion", padding='left')

prompt_format = """
# Here's a yaml file to offer a completion for
# Lines after the current one
{text_after}
# Lines before the current one
{text_before}
# Completion:
"""

input_prefix = """
name: my_awesome_env
dependencies:

"""

input_suffix = ""

generator = pipeline('text-generation', model=model, tokenizer=tokenizer, device='cuda')

generator(prompt_format.format(text_after=input_suffix, text_before=input_prefix), max_new_tokens=64)

# [{'generated_text': "\n# Here's a yaml file to offer a completion for\n# Lines after the current one\n\n# Lines before the current one\n\nname: my_awesome_env\ndependencies:\n\n\n# Completion:\n- deploy"}]