Zamba-7B-v1 is a hybrid model between Mamba, a state-space model, and transformers. It uses a mamba backbone with a shared transformer layer every 6 blocks. Zamba was trained using next-token prediction. It uses the Mistral v0.1 tokenizer. We came to this architecture after a series of ablations at small scales. Zamba-7B-v1 was pre-trained on 1T tokens of text and code data sourced from open web-datasets. Subsequently in a second phase, Zamba was annealed on a mixture of 50B high-quality tokens.
Note: the current Huggingface implementation of Zamba performs slower than our internal implementation. We are working to fix this with the Huggingface team.
Our technical report describing the training of Zamba is available here.
Quick start
Presequities
To download Zamba, clone Zyphra's fork of transformers:
In order to run optimized Mamba implementations on a CUDA device, you need to install mamba-ssm and causal-conv1d:
pip install mamba-ssm causal-conv1d>=1.2.0
You can run the model without using the optimized Mamba kernels, but it is not recommended as it will result in significantly higher latency.
To run on CPU, please specify use_mamba_kernels=False when loading the model using AutoModelForCausalLM.from_pretrained.
Inference
python
1from transformers import AutoTokenizer, AutoModelForCausalLM
2import torch
34tokenizer = AutoTokenizer.from_pretrained("Zyphra/Zamba-7B-v1")5model = AutoModelForCausalLM.from_pretrained("Zyphra/Zamba-7B-v1", device_map="auto", torch_dtype=torch.bfloat16)67input_text ="What factors contributed to the fall of the Roman Empire?"8input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")910outputs = model.generate(**input_ids, max_new_tokens=100)11print(tokenizer.decode(outputs[0]))
To load a different checkpoint use, e.g., for iteration 2500,
model = AutoModelForCausalLM.from_pretrained("Zyphra/Zamba-7B-v1", device_map="auto", torch_dtype=torch.bfloat16, revision="iter2500")
The default iteration is the fully trained model, corresponding to iteration 25156. This is the number of training iterations done starting from Zamba-phase 1 Zyphra/Zamba-7B-v1-phase1. See arXiv:2405.16712 for more details on training.
Model Details
Zamba utilizes a unique hybrid SSM architecture. This architecture consists of a backbone of Mamba layers interspersed with a shared attention layer. This attention has shared weights to minimize the parameter cost of the model. We find that concatenating the original model embeddings to the input to this attention block improves performance, likely due to better maintenance of information across depth.
Performance
We find that Zamba performs significantly better than existing open models (with open datasets and training details) at this scale. However, it performs slightly worse than the leading open-weight models at the 7B scale. Most of this difference derives from MMLU and reasoning evaluations. Zamba, however, is trained on significantly fewer tokens than these models and is the most sample efficient model in terms of performance per training tokens.
Due to its SSM architecture, Zamba is extremely efficient in inference, substantially outperforming comparable 7B and 8B models in inference latency as well as memory cost of generation due to its substantially diminished KV cache.
Citation
If you find Zamba useful in your work please cite it as:
@article{glorioso2024zamba,
title={Zamba: A Compact 7B SSM Hybrid Model},
author={Glorioso, Paolo and Anthony, Quentin and Tokpanov, Yury and Whittington, James and Pilault, Jonathan and Ibrahim, Adam and Millidge, Beren},
journal={arXiv preprint arXiv:2405.16712},
year={2024}
}
Notice
Zamba is a pretrained base model and therefore does not have any moderation mechanism. In addition, one should not expect good chat performance, as this model was not fine-tuned for chat.
