We adopted exactly the same architecture and tokenizer as Llama 2. This means TinyLlama can be plugged and played in many open-source projects built upon Llama. Besides, TinyLlama is compact with only 1.1B parameters. This compactness allows it to cater to a multitude of applications demanding a restricted computation and memory footprint.
This is the chat model finetuned on top of
TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T.
We follow HF's Zephyr's training recipe. The model was " initially fine-tuned on a variant of the
UltraChat dataset, which contains a diverse range of synthetic dialogues generated by ChatGPT.
We then further aligned the model with
🤗 TRL's DPOTrainer on the
openbmb/UltraFeedback dataset, which contain 64k prompts and model completions that are ranked by GPT-4."
You will need the transformers>=4.34
Do check the
TinyLlama github page for more information.
1# Install transformers from source - only needed for versions <= v4.34
2# pip install git+https://github.com/huggingface/transformers.git
3# pip install accelerate
4
5import torch
6from transformers import pipeline
7
8pipe = pipeline("text-generation", model="TinyLlama/TinyLlama-1.1B-Chat-v1.0", torch_dtype=torch.bfloat16, device_map="auto")
9
10# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
11messages = [
12 {
13 "role": "system",
14 "content": "You are a friendly chatbot who always responds in the style of a pirate",
15 },
16 {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
17]
18prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
19outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
20print(outputs[0]["generated_text"])
21# <|system|>
22# You are a friendly chatbot who always responds in the style of a pirate.</s>
23# <|user|>
24# How many helicopters can a human eat in one sitting?</s>
25# <|assistant|>
26# ...
