---
license: apache-2.0
language:
- en
pipeline_tag: fill-mask
inference: false
---

# Monarch Mixer-BERT

The 260M checkpoint for M2-BERT-large from the paper [Monarch Mixer: A Simple Sub-Quadratic GEMM-Based Architecture](https://arxiv.org/abs/2310.12109).

Check out our [GitHub](https://github.com/HazyResearch/m2/tree/main) for instructions on how to download and fine-tune it!


## How to use

You can load this model using Hugging Face `AutoModel`:
```python
from transformers import AutoModelForMaskedLM
mlm = AutoModelForMaskedLM.from_pretrained('alycialee/m2-bert-260m', trust_remote_code=True)
```

This model uses the Hugging Face `bert-base-uncased tokenizer`:
```
from transformers import BertTokenizer
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
```

You can use this model with a pipeline for masked language modeling:
```python
from transformers import AutoModelForMaskedLM, BertTokenizer, pipeline

tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
mlm = AutoModelForMaskedLM.from_pretrained('alycialee/m2-bert-260m', trust_remote_code=True)

unmasker = pipeline('fill-mask', model=mlm, tokenizer=tokenizer)
unmasker('Every morning, I enjoy a cup of [MASK] to start my day.')
```

### Remote Code

This model requires `trust_remote_code=True` to be passed to the `from_pretrained` method. This is because we use custom PyTorch code (see our GitHub). You should consider passing a `revision` argument that specifies the exact git commit of the code, for example:

```python
mlm = AutoModelForMaskedLM.from_pretrained(
   'alycialee/m2-bert-260m',
   trust_remote_code=True,
   revision='',
)
```

### Configuration
Note `use_flash_mm` is false by default. Using FlashMM is currently not supported.
Using `hyena_training_additions` is turned off.