README.md

metadata

widget:
  - text: 'The meaning of life is to '
datasets:
  - tatsu-lab/alpaca
  - ehartford/ultrachat-uncensored
  - codeparrot/github-code
language:
  - en
library_name: transformers
pipeline_tag: conversational

Dromedary - 7B

Dromedary is our uncensored flagship model, designed for programming tasks and communication. Dromedary is a fine-tune of LLAMA-2. Dromedary was fine-tuned on 3 public datasets and on 1 private dataset (synthetic, by our model GPT-LIO.), ranging from programming to healthcare advice.

This model supports both chat and text completion

Technical Information

Dromedary is a model designed to be unbiased and uncensored. However, this model was detoxified. (we dont want any bad people!!!) The model was tuned with two RTX 8000s, at 1 epoch.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("intone/Dromedary-7B")
model = AutoModelForCausalLM.from_pretrained("intone/Dromedary-7B", device_map="auto", torch_dtype='auto')

messages = [
    {"role": "user", "content": "Hi, how are you?"}
]

input_ids = tokenizer.apply_chat_template(conversation=messages, tokenize=True, add_generation_prompt=True, return_tensors='pt')
output_ids = model.generate(input_ids.to('cuda'))
response = tokenizer.decode(output_ids[0][input_ids.shape[1]:], skip_special_tokens=True)

print(response) # --> "Hello! How can I assist you today?"