README.md

---
library_name: transformers
tags: []
---

# ***Mol-MoE***: Training Preference-Guided Routers for Molecule Generation
*Diego Calanzone (1, 2), Pierluca D'Oro (2), Pierre-Luc Bacon (1, 2)* <br>
*(1) Universite de Montreal, (2) Mila Quebec AI Institute* <br>
**arXiv**: https://arxiv.org/abs/2502.05633

**Abstract**: Recent advances in language models have enabled framing molecule generation as sequence modeling. However, existing approaches often rely on single-objective reinforcement learning, limiting their applicability to real-world drug design, where multiple competing properties must be optimized. Traditional multi-objective reinforcement learning (MORL) methods require costly retraining for each new objective combination, making rapid exploration of trade-offs impractical. To overcome these limitations, we introduce Mol-MoE, a mixture-of-experts (MoE) architecture that enables efficient test-time steering of molecule generation without retraining. Central to our approach is a preference-based router training objective that incentivizes the router to combine experts in a way that aligns with user-specified trade-offs. This provides improved flexibility in exploring the chemical property space at test time, facilitating rapid trade-off exploration. Benchmarking against state-of-the-art methods, we show that Mol-MoE achieves superior sample quality and steerability.


## How to use this model
This LM is fine-tuned to generate molecules in the SMILES format wrt. desired properties.
For unconditioned SMILES generation, use the BOS token `<s>`. <br>
```
prompt: <s> 
```

An example of the generation pipeline:
```
from transformers import AutoTokenizer, AutoModelForCausalLM
import re

# Setup
device = "cuda"
tokenizer = AutoTokenizer.from_pretrained("ddidacus/MORLHF-mol-llama-1b")
model = AutoModelForCausalLM.from_pretrained("ddidacus/MORLHF-mol-llama-1b")
generation_kwargs = {
    "max_new_tokens": 128,
    "min_length": -1,
    "top_k": 0.0,
    "top_p": 0.9,
    "do_sample": True,
    "pad_token_id": tokenizer.eos_token_id,
    "temperature": 1.0
}

# Inference
query = "<JNK3=0.3><DRD2=0.7><GSK3B=0.2><CYP2D6=0.8><CYP2C19=0.8><s>"
toks = tokenizer([query], return_tensors="pt")["input_ids"].to(device)
output = model.generate(toks, **generation_kwargs)
output = tokenizer.batch_decode(output)

# Parsing
filter = r'<s>(.*?)</s>'
molecule = re.findall(filter, output[0], re.DOTALL)
```

### Model Description
This model is a fine-tuned version of LLaMa 3.2 1B through two stages:
1. Fine-tuning on ~3.5M molecules extracted from: ZINC 250K, MOSES, CHEMBL
2. RLHF-tuning using instruction fine-tuning on 5 distinct reward signals.

The detailed pipeline we followed is reported in the original paper: <br>
*"Rewards-in-Context: Multi-objective Alignment of Foundation Models with Dynamic Preference Adjustment" Yang et al. 2024* [1]

- **Developed by:** Diego Calanzone ([email protected])
- **Model type:** Decoder Only Transformer
- **Finetuned from model [optional]:** LLaMA 3.2 1B

Read the paper for further details.

### Sources
[1] https://arxiv.org/abs/2402.10207

<!--
### Model Sources [optional]

- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]

## Uses

### Direct Use

[More Information Needed]

### Downstream Use [optional]

[More Information Needed]

### Out-of-Scope Use

[More Information Needed]

## Bias, Risks, and Limitations

[More Information Needed]

### Recommendations
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

## How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

## Training Details

### Training Data

[More Information Needed]

### Training Procedure

#### Preprocessing [optional]

[More Information Needed]


#### Training Hyperparameters

- **Training regime:** [More Information Needed] 

#### Speeds, Sizes, Times [optional]

[More Information Needed]

## Evaluation

### Testing Data, Factors & Metrics

#### Testing Data

[More Information Needed]

#### Factors

[More Information Needed]

#### Metrics

[More Information Needed]

### Results

[More Information Needed]

#### Summary



## Model Examination [optional]

[More Information Needed]

## Environmental Impact

Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).

- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]

## Technical Specifications [optional]

### Model Architecture and Objective

[More Information Needed]

### Compute Infrastructure

[More Information Needed]

#### Hardware

[More Information Needed]

#### Software

[More Information Needed]

## Citation [optional]

**BibTeX:**

[More Information Needed]

**APA:**

[More Information Needed]

## Glossary [optional]

[More Information Needed]

## More Information [optional]

[More Information Needed]

## Model Card Authors [optional]

[More Information Needed]

## Model Card Contact

[More Information Needed] -->