Downloads · 30 days
0
slm-stories
hyrajratan/slm-stories
slm-stories is a text generation model from hyrajratan. Use it when you need the model to write or continue text. The card lists the license as mit.
A compact GPT-style transformer trained on the TinyStories dataset to generate short children's stories.
Access
Public
Updated May 12, 2026
Repo size
619 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt619 MB · 100%
At a glance
- Task
- Text Generation
- License
- mit
- Access
- Public
- Created
- May 12, 2026
- Updated
- May 12, 2026
- SHA
- d487c806
Try a prompt
Model card
From the Hugging Face model README
SLM Stories -- Small Language Model
A compact GPT-style transformer trained on the TinyStories dataset to generate short children's stories.
Architecture
| Parameter | Value |
|---|---|
| Layers | 8 transformer blocks |
| Hidden dim | 512 |
| Attention heads | 8 |
| Context length | 256 tokens |
| Vocabulary | GPT-2 BPE (50,257 tokens) |
| Activation | SwiGLU |
| Normalization | RMSNorm |
Usage
import torch, tiktoken
from huggingface_hub import hf_hub_download
from config import SLMConfig
from model import SLM
from generate import generate_story
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
ckpt = torch.load(hf_hub_download("hyrajratan/slm-stories", "best.pt"), map_location=device)
cfg = SLMConfig()
model = SLM(cfg).to(device)
model.load_state_dict({k.removeprefix("_orig_mod."): v for k, v in ckpt["model_state"].items()})
model.eval()
enc = tiktoken.get_encoding("gpt2")
print(generate_story(model, enc, cfg, device, "Once upon a time there was a little cat"))
Live Demo
Deployed on Streamlit Cloud.
- Task
- Text Generation
- License
- mit
- Languages
- en
- Created
- May 12, 2026
- Updated
- May 12, 2026