Instructions to use purrgpt-community/Tiny-Purr-1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use purrgpt-community/Tiny-Purr-1B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="purrgpt-community/Tiny-Purr-1B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("purrgpt-community/Tiny-Purr-1B")
model = AutoModelForCausalLM.from_pretrained("purrgpt-community/Tiny-Purr-1B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use purrgpt-community/Tiny-Purr-1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "purrgpt-community/Tiny-Purr-1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "purrgpt-community/Tiny-Purr-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/purrgpt-community/Tiny-Purr-1B

SGLang

How to use purrgpt-community/Tiny-Purr-1B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "purrgpt-community/Tiny-Purr-1B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "purrgpt-community/Tiny-Purr-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "purrgpt-community/Tiny-Purr-1B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "purrgpt-community/Tiny-Purr-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use purrgpt-community/Tiny-Purr-1B with Docker Model Runner:
```
docker model run hf.co/purrgpt-community/Tiny-Purr-1B
```

🐾 Tiny-Purr-1B Model Card

Model Name: Tiny-Purr-1B Base Model: LiquidAI/LFM2-1.2B Architecture: Causal Language Model (Decoder-only Transformer) Fine-tuning Method: LoRA (Low-Rank Adaptation)

🧠 Model Summary

Tiny-Purr-1B is a playful, cat-themed AI assistant trained via LoRA fine-tuning on the Claude_32K conversational dataset. The model adopts a lighthearted, sarcastic, and friendly personality — like a talkative cat that helps users with humor and a paw-sitive attitude.

The base model, LiquidAI/LFM2-1.2B, provides a strong multilingual and instruction-following foundation. Tiny-Purr-1B enhances this with personality-rich responses and conversational alignment through dialogue-style fine-tuning.

🧩 Training Overview

Aspect	Details
Base Model	LiquidAI/LFM2-1.2B
Dataset	SicariusSicariiStuff/Claude_32K
Training Objective	Causal language modeling (next-token prediction)
Fine-tuning Method	LoRA (Low-Rank Adaptation)
Precision	4-bit base model loading + FP16 training
Hardware	NVIDIA T4 GPU
Training Steps	~2000 samples × 3 epochs
Optimizer	AdamW (via Hugging Face `Trainer`)
Batch Size	1 (per device) with gradient accumulation (16 steps)
Learning Rate	5e-4
Context Length	2048 tokens
LoRA Parameters	`r=8`, `alpha=16`, `dropout=0.05` on `q_proj` and `v_proj`

💬 Conversational Formatting

The dataset is reformatted with special tokens for multi-turn conversations, modeled after instruction-tuned chat systems. Each conversation is wrapped with:

<|startoftext|>
<|im_start|>system
You are Tiny-Purr, a friendly, sarcastic, playful AI assistant in the form of a cat developed by PurrGPT Community...
<|im_start|>user
User message...
<|im_end|>
<|im_start|>assistant
Model reply...
<|im_end|>

This ensures proper role conditioning and consistent generation in chat-style contexts.

🧪 Evaluation & Intended Use

Intended Use

Chat-based assistant with personality-rich, fun, and casual dialogue.
Creative writing and character roleplay.
Non-serious Q&A and social interactions.

Not Intended For

Critical decision-making (e.g., medical, legal, or financial).
Factual retrieval where precision is required.
Harmful or explicit content generation.

Qualitative Observations

During spot evaluations, Tiny-Purr-1B:

Demonstrates humorous and playful language.
Maintains consistent “cat persona” tone.
Handles multi-turn conversation gracefully (within 2K token limit).

🧰 Technical Details

Special Tokens Added:

<|startoftext|>, <|im_start|>system, <|im_start|>user,
<|im_start|>assistant, <|im_end|>

Merging Process:

Trained LoRA adapters merged into base model using merge_and_unload().
Tokenizer and chat templates copied from base repository for deployment readiness.

⚠️ Limitations

Context length limited to 2048 tokens; longer conversations may truncate.
Humor bias: Playful tone may interfere with formal or factual responses.
Not safety-aligned: No RLHF or safety-tuned layers applied beyond dataset filtering.
Possible hallucinations due to base model behavior.

🧾 Acknowledgements

Base model: LiquidAI/LFM2-1.2B
Dataset: Claude_32K by SicariusSicariiStuff

Downloads last month: 12

Safetensors

Model size

1B params

Tensor type

F32

F16

Model tree for purrgpt-community/Tiny-Purr-1B

Base model

LiquidAI/LFM2-1.2B

Adapter

(7)

this model

Space using purrgpt-community/Tiny-Purr-1B 1

Collection including purrgpt-community/Tiny-Purr-1B

Tiny-Purr

Collection

Tiny-Purr, a OSS clone of PurrGPT! • 4 items • Updated Mar 2