---

license: apache-2.0

language:

• en

base_model:

• mondk/Msh-Tiny-47M

pipeline_tag: text-generation

tags:

• gguf
• from-scratch
• msh
• f16
• q2k
• q4km
• ty

datasets:

• mondk/Greetings-hi-for-train-Msh-v2
• tatsu-lab/alpaca
• databricks/databricks-dolly-15k
• teknium/OpenHermes-2.5
• m-a-p/Code-Feedback
• OpenAssistant/oasst1
• FreedomIntelligence/medical-o1-reasoning-SFT
• glaiveai/glaive-function-calling-v2
• openai/openai_humaneval
• HuggingFaceH4/no_robots
• open-thoughts/OpenThoughts-114k
• HuggingFaceH4/ultrachat_200k
• google-research-datasets/poem_sentiment
• CohereLabs/aya_dataset
• sentence-transformers/natural-questions

---

msh-tiny (GGUF)

A tiny (~47M parameter) GPT-2-style chat model, trained completely from scratch — no pretrained base model, custom BPE tokenizer trained from zero, custom PyTorch transformer architecture. This repo contains GGUF builds for use with llama.cpp, Ollama, and LM Studio.

The .safetensors source model is at mondk/Msh-Tiny-47M.

Limitations

This model was trained from random initialization on a modest amount of data with limited compute — it is a small educational project, not a production-quality assistant. Expect it to follow the chat format reliably but produce limited/inconsistent knowledge and occasional incoherent answers.

Prompt format

<|user|>
{your message}
<|assistant|>

The model was trained to stop generating at <|end|>.

Usage

Ollama

FROM ./model-f16.gguf

ollama create msh-tiny -f Modelfile
ollama run msh-tiny

LM Studio: drop the .gguf file into your models folder and load it directly.

llama.cpp

./llama-cli -m model-f16.gguf -p "<|user|>\nhi\n<|assistant|>\n"

Training data

Combining 15 well-known open instruction/chat datasets plus a small hand-written set of everyday chit-chat (greetings, thanks, small talk):

• mondk/Greetings-hi-for-train-Msh-v2
• tatsu-lab/alpaca
• databricks/databricks-dolly-15k
• teknium/OpenHermes-2.5
• m-a-p/Code-Feedback
• OpenAssistant/oasst1
• FreedomIntelligence/medical-o1-reasoning-SFT
• glaiveai/glaive-function-calling-v2
• openai/openai_humaneval
• HuggingFaceH4/no_robots
• open-thoughts/OpenThoughts-114k
• HuggingFaceH4/ultrachat_200k
• google-research-datasets/poem_sentiment
• CohereLabs/aya_dataset
• sentence-transformers/natural-questions