metadata

library_name: transformers
license: mit
datasets:
  - MattCoddity/dockerNLcommands
language:
  - en
old_version: Bruce1489/Llama3.2-docker-command-v2

Model Card for Model ID

Improved accuracy over V2.

Fine-tuned the "meta-llama/Llama-3.2-1B-Instruct" model using the QLoRA technique.

The resulting model can generate Docker commands based on given statements.

Developed by: Bruce1489(정성훈, Bruce Jung)
License: mit
Finetuned from model [optional]: meta-llama/Llama-3.2-1B-Instruct

Direct Use

import torch
from transformers import pipeline

pipe = pipeline(
    "text-generation",
    model="Bruce1489/Llama3.2-docker-command-v3",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
messages = [
    {"role": "system", "content": "translate this sentence in docker command"},
    {"role": "user", "content": "Please show me the Docker containers that have exited and are related to the mongo image."},
]

outputs = pipe(
    messages,
    max_new_tokens=256,
    temperature = 0.5,
    top_p  = 0.9,
    top_k = 10,
    do_sample = True,
    repetition_penalty = 1.2
)
print(outputs[0]["generated_text"][-1]['content'])