34 lines
948 B
Markdown
34 lines
948 B
Markdown
---
|
|
license: apache-2.0
|
|
base_model: Qwen/Qwen3-1.7B
|
|
tags:
|
|
- biology
|
|
- bio-posttrain
|
|
- dna-rl
|
|
- dna
|
|
library_name: transformers
|
|
---
|
|
|
|
# Bio-posttrain Qwen3-1.7B DNA RL
|
|
|
|
DNA reinforcement learning (GRPO) 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`
|
|
- **DNA encoder:** Evo2 `evo2_1b_base` (frozen; not included)
|
|
- **Embedding layer:** `blocks.20.mlp.l3`
|
|
- **SFT pool:** sft256 / LoRA rank 16
|
|
- **GRPO LoRA:** rank 16, alpha 32
|
|
- **Checkpoint:** step 1156
|
|
|
|
This repo contains the merged text LLM plus `dna_projection.pt`.
|
|
|
|
## Loading
|
|
|
|
Same layout as the DNA SFT model — see [bio-posttrain-qwen3-1.7b-dna-sft](https://huggingface.co/mims-harvard/bio-posttrain-qwen3-1.7b-dna-sft).
|
|
|
|
## Collection
|
|
|
|
Part of the [Bio-posttrain](https://huggingface.co/collections/mims-harvard/bio-posttrain) collection.
|