---
license: apache-2.0
base_model: ByteDance/Ouro-1.4B-Thinking
library_name: transformers
pipeline_tag: text-generation
tags:
  - looped-language-model
  - recurrent-depth
  - reinforcement-learning
  - grpo
  - math
---

# Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

REPO = "omar81939/Ouro-1.4B-Thinking-depth-GRPO"
DEPTH = 16

tokenizer = AutoTokenizer.from_pretrained(REPO)
model = AutoModelForCausalLM.from_pretrained(
    REPO,
    trust_remote_code=True,
    dtype="bfloat16",
    total_ut_steps=DEPTH,
)
```

With vLLM, set `hf_overrides={"total_ut_steps": DEPTH}` when creating the engine.