Instruction-Tuned Language Model

This repository contains an instruction-tuned causal language model for text generation and chat-style prompts.

Model details

• Type: causal language model
• Parameters: approximately 1.54 billion
• Non-embedding parameters: approximately 1.31 billion
• Layers: 28
• Architecture: rotary position embeddings, gated feed-forward layers, RMS normalization, attention query/key/value bias, and tied word embeddings
• Attention: grouped-query attention with 12 query heads and 2 key/value heads
• Context length: 32,768 tokens, with generation up to 8,192 tokens
• Languages: multilingual text generation

Requirements

Use a recent version of transformers that supports this model architecture.

Quickstart

The following example loads the model from this repository and generates a response.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "latent-artist/66c200392ff148279a5995ef1455d3de"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Give me a short introduction to large language models."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]