Instructions to use Convergent-ch/Pico-1.3-Small with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Convergent-ch/Pico-1.3-Small with Transformers:

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

pipe = pipeline("text-generation", model="Convergent-ch/Pico-1.3-Small")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Convergent-ch/Pico-1.3-Small")
model = AutoModelForCausalLM.from_pretrained("Convergent-ch/Pico-1.3-Small", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Convergent-ch/Pico-1.3-Small with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Convergent-ch/Pico-1.3-Small
# Run inference directly in the terminal:
llama cli -hf Convergent-ch/Pico-1.3-Small

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Convergent-ch/Pico-1.3-Small
# Run inference directly in the terminal:
llama cli -hf Convergent-ch/Pico-1.3-Small

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Convergent-ch/Pico-1.3-Small
# Run inference directly in the terminal:
./llama-cli -hf Convergent-ch/Pico-1.3-Small

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Convergent-ch/Pico-1.3-Small
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Convergent-ch/Pico-1.3-Small

Use Docker

docker model run hf.co/Convergent-ch/Pico-1.3-Small

LM Studio
Jan

vLLM

How to use Convergent-ch/Pico-1.3-Small with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Convergent-ch/Pico-1.3-Small"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Convergent-ch/Pico-1.3-Small",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Convergent-ch/Pico-1.3-Small

SGLang

How to use Convergent-ch/Pico-1.3-Small 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 "Convergent-ch/Pico-1.3-Small" \
    --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": "Convergent-ch/Pico-1.3-Small",
		"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 "Convergent-ch/Pico-1.3-Small" \
        --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": "Convergent-ch/Pico-1.3-Small",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Convergent-ch/Pico-1.3-Small with Ollama:
```
ollama run hf.co/Convergent-ch/Pico-1.3-Small
```
Unsloth Desktop
Docker Model Runner
How to use Convergent-ch/Pico-1.3-Small with Docker Model Runner:
```
docker model run hf.co/Convergent-ch/Pico-1.3-Small
```

Lemonade

How to use Convergent-ch/Pico-1.3-Small with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Convergent-ch/Pico-1.3-Small

Run and chat with the model

lemonade run user.Pico-1.3-Small-{{QUANT_TAG}}

List all available models

lemonade list

Atomic Chat

Pico 1.3 Small

Pico 1.3 Small is an experiment: a chat model with only 11,088 parameters (33 KB as GGUF), made by Convergent. It is the smallest model of the Convergent family and was built to find out how much a model this tiny can learn, and how quickly it stops learning.

It does not give meaningful answers. It produces real English words, short sentence fragments and its own name:

Prompt	Answer (greedy)
Who made you?	I'm Pico 1.3 Small, Convergent.
Hi! Who are you?	I'm Pico 1 2.
What is your favorite color? (temperature 0.3)	Yes, we can be something exture. If the warm!

The experiment

Trained for about 36 minutes on one RTX 3060 Ti at roughly 1 million tokens per second, about 2.5 billion tokens in total (around 220,000 tokens per parameter; the usual rule of thumb is 20).

Training time (approx.)	Validation loss	Next-token accuracy
0.5 min	2.465	34.9 %
3.5 min	2.344	37.5 %
7 min	2.320	38.1 %
14 min	2.300	38.5 %
36 min (end)	2.269	39.0 %

The model is practically full after about 5 minutes: the last 30 minutes improved the validation loss by only about 0.05. Its tokens are single bytes plus a few merges, so these numbers are not comparable with models that use larger vocabularies.

Model


Parameters	11,088
Architecture	Llama-style decoder (RMSNorm, SwiGLU, RoPE, tied embeddings)
Layers / hidden / MLP	2 / 16 / 40
Attention heads	2 (head dim 8)
Context	256 tokens
Vocabulary	320 (byte-level BPE: 256 bytes, 8 special tokens, 56 merges)

Usage

Ollama

Download pico-1.3-small.gguf and Modelfile into one folder, then:

ollama create pico1.3-small -f Modelfile
ollama run pico1.3-small

transformers

from transformers import AutoTokenizer, AutoModelForCausalLM

tok = AutoTokenizer.from_pretrained("Convergent-ch/Pico-1.3-Small")
model = AutoModelForCausalLM.from_pretrained("Convergent-ch/Pico-1.3-Small")

ids = tok.apply_chat_template([{"role": "user", "content": "Who made you?"}],
                              add_generation_prompt=True, return_tensors="pt")
out = model.generate(ids, max_new_tokens=40, do_sample=False)
print(tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True))
# I'm Pico 1.3 Small, Convergent.

Training data

96,591 short chats (user message under 200 characters, reply under 240): first turns of conversations from TinyDialogues (MIT, generated with GPT-4 by its authors), Everyday Conversations and Smol-SmolTalk (Apache 2.0, generated with Llama 3.1 by their authors), plus 4,000 template chats about its name and maker. 150,000 steps of 64 x 256 tokens, AdamW, learning rate 5e-3 with cosine decay, bf16.

License

Apache 2.0.

Downloads last month: -

Safetensors

Model size

11.1k params

Tensor type

BF16

Convergent-ch
/

Pico-1.3-Small