nishantup/nanogpt-slm-124m

nanoGPT SLM -- 123.8M Parameter Language Model

A small language model trained from scratch using a custom nanoGPT implementation. Pretrained on 133 classic English fiction novels from Project Gutenberg.

Quick Start

Option 1: Run directly (downloads model + runs examples)

pip install torch tiktoken huggingface_hub
python nanogpt_slm_pretrained_inference.py

Option 2: Import and use ask() in your own code

# Import loads the model automatically (one-time download from HuggingFace)
## nanogpt_slm_pretrained_inference.py

from nanogpt_slm_pretrained_inference import ask, generate_text

# Text completion
print(ask("Once upon a time there was"))
print()

# Control generation
print(ask(
    "The meaning of life is",
    temperature=1.0,    # higher = more creative
    top_k=100,          # wider sampling pool
    max_tokens=150      # longer output
))
print()

# generate_text is an alias for ask
print(generate_text("She opened the door and saw", max_tokens=200))
print()

Option 3: Load weights manually

from huggingface_hub import hf_hub_download
import torch

model_path = hf_hub_download(
    repo_id="nishantup/nanogpt-slm-124m",
    filename="nanogpt_slm_best.pth"
)

# Full architecture in nanogpt_slm_pretrained_inference.py
from nanogpt_slm_pretrained_inference import GPT, GPTKV, GPTConfig

config = GPTConfig()
model = GPTKV(config)  # KV-cache enabled variant
model.load_state_dict(torch.load(model_path, map_location="cpu"))
model.eval()

Model Details

Attribute	Value
Parameters	123.8M
Architecture	nanoGPT (12 layers, 12 heads, 768 dim)
Context length	256 tokens
Tokenizer	tiktoken GPT-2 BPE (50,257 tokens)
Training data	133 English fiction novels (37.5M tokens)
Framework	PyTorch

Files

File	Description
nanogpt_slm_best.pth	Pretrained model weights (best val loss)
nanogpt_slm_pretrained_inference.py	Standalone inference script -- import and call ask()
config.json	Model configuration

ask() / generate_text() API Reference

ask(prompt, max_tokens=200, temperature=0.8, top_k=40)
generate_text(prompt, max_tokens=200, temperature=0.8, top_k=40)  # alias

Parameter	Default	Description
prompt	(required)	Text to continue from
max_tokens	200	Maximum tokens to generate
temperature	0.8	0.01 = near-greedy, 0.8 = balanced, 1.5 = creative
top_k	40	Top-k filtering (None = no filtering)

Fine-tuned Variants

Variant	Repo
Instruction-tuned (SFT)	nishantup/nanogpt-slm-instruct

Notes

• Trained completely from scratch (no pretrained initialization)
• Uses KV cache (GPTKV) for O(1) per-token decode
• Weight tying between token embeddings and LM head