---
license: apache-2.0
base_model: Qwen/Qwen3-1.7B
tags:
  - biology
  - bio-posttrain
  - rna-sft
  - rna
library_name: transformers
---

# Bio-posttrain Qwen3-1.7B RNA SFT

RNA supervised fine-tuning (SFT) checkpoint from [How Post-Training Shapes Biological Reasoning Models](https://huggingface.co/collections/mims-harvard/bio-posttrain).

## Model details

- **Base model:** `Qwen/Qwen3-1.7B`
- **RNA embeddings:** Precomputed TranscriptFormer (2048-d), projected via `rna_projection.pt`
- **LoRA:** rank 32, alpha 64
- **Validation loss:** 0.5394

## Files

| File | Description |
|------|-------------|
| `model.safetensors` | Merged LLM weights |
| `rna_projection.pt` | Linear map RNA embed (2048) → text hidden |
| `rna_model_config.json` | Architecture metadata |

## Loading

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("mims-harvard/bio-posttrain-qwen3-1.7b-rna-sft", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("mims-harvard/bio-posttrain-qwen3-1.7b-rna-sft")
proj = torch.load("rna_projection.pt", map_location="cpu")
```

RNA sequence embeddings are supplied offline at inference time. See the `rna_models` code in the BioReason repository.

## Collection

Part of the [Bio-posttrain](https://huggingface.co/collections/mims-harvard/bio-posttrain) collection.