---
library_name: transformers
pipeline_tag: text-generation
base_model:
- LiquidAI/LFM3-1.7B
---

This is a randomly initialized tiny model based on [LiquidAI/LFM3-1.7B](https://huggingface.co/LiquidAI/LFM3-1.7B) for testing and debugging.

| Feature | Source | This repo |
|---|---|---|
| Checkpoint tensor elements | 1.73B BF16 | 8.62M BF16 |
| Safetensors weight files | 3.46 GB (model.safetensors) | 17.26 MB (model.safetensors) |
| Layers / layer-type cycle | 24 / two-conv prefix, then full/conv/sliding/conv | 10 / same prefix and two complete cycles |
| Hidden width / MLP width | 2,048 / 7,680 | 64 / 128 |
| Q heads / KV heads / head_dim | 16 / 4 / 128 | 4 / 1 / 32 |
| Short-conv kernel / cache | 7 / 7 | Same |
| RoPE | full 0.5 partial, theta 1e6; sliding theta 1e4 | Same |
| Sandwich norms / QK RMSNorm / sigmoid head gate | present | Same |
| Vocabulary rows / embedding tying | 128,000 / tied | Same |
| Maximum context length | 262,144 | 128,000 |
| Stored precision | BF16 | Same |

Residual width 64 and MLP width 128 follow [tiny-random/lfm2](https://huggingface.co/tiny-random/lfm2). The pinned upstream implementation is included for Transformers loading. This repo provides a Transformers example; no serving-backend validation is claimed.

| File path | Size |
|------|------|
| model.safetensors | 17.3MB |

### Example usage

#### Transformers (✅ verified)

Load the included pinned upstream implementation with `trust_remote_code=True`.

```python
def run_transformers(model_id: str):
    import torch
    from transformers import AutoModelForCausalLM, AutoTokenizer

    model_id = str(model_id)
    tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
    model = (
        AutoModelForCausalLM.from_pretrained(
            model_id, trust_remote_code=True, dtype=torch.bfloat16, attn_implementation="sdpa"
        )
        .eval()
        .cuda()
    )
    inputs = tokenizer.apply_chat_template(
        [{"role": "user", "content": "What is C. elegans?"}],
        add_generation_prompt=True,
        return_tensors="pt",
        return_dict=True,
        enable_thinking=False,
    ).to(model.device)
    with torch.no_grad():
        logits = model(**inputs).logits
        assert torch.isfinite(logits).all()
        output = model.generate(**inputs, do_sample=False, max_new_tokens=32)
    print(tokenizer.decode(output[0, inputs["input_ids"].shape[1] :], skip_special_tokens=False))


if __name__ == "__main__":
    run_transformers(model_id="tiny-random/lfm3")
```

### Creating this repo

- Code: [.meta/create_model.py](.meta/create_model.py)
- Model structure: [.meta/model.log](.meta/model.log)
- Environment: [.meta/env.json](.meta/env.json)

### Change log

- 2026-10-30: Initial version.
- 2026-10-31: Resize to hidden width 64 and MLP width 128, following tiny-random/lfm2; use four Q heads and one KV head while preserving the LFM3 layer cadence.

*This repo is generated by GPT-6.1-Sol.*