Find below some example scripts on how to use the model in transformers (Make sure to have the latest transformers, or the one built from source):
Using the Pytorch model
Running the model on a CPU
Click to expand
python
1from transformers import AutoTokenizer, AutoModelForCausalLM
23tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-mamba-7b-instruct")4model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-mamba-7b-instruct")56# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating7messages =[8{"role":"user","content":"How many helicopters can a human eat in one sitting?"},9]1011input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)12input_ids = tokenizer(input_text, return_tensors="pt").input_ids
1314outputs = model.generate(input_ids, max_new_tokens=30)15print(tokenizer.decode(outputs[0]))
Running the model on a GPU
Click to expand
python
1# pip install accelerate2from transformers import AutoTokenizer, AutoModelForCausalLM
34tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-mamba-7b-instruct")5model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-mamba-7b-instruct", device_map="auto")67# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating8messages =[9{"role":"user","content":"How many helicopters can a human eat in one sitting?"},10]1112input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)13input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")1415outputs = model.generate(input_ids, max_new_tokens=30)16print(tokenizer.decode(outputs[0]))
Running the model on a GPU using torch.compile
Click to expand
python
1import torch
2from transformers import AutoTokenizer, AutoModelForCausalLM
34tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-mamba-7b-instruct")5model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-mamba-7b-instruct", torch_dtype=torch.bfloat16).to(0)67model = torch.compile(model)89# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating10messages =[11{"role":"user","content":"How many helicopters can a human eat in one sitting?"},12]13input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)14input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")1516outputs = model.generate(input_ids, max_new_tokens=30)17print(tokenizer.decode(outputs[0]))
Running the model on a GPU using different precisions
FP16
Click to expand
python
1# pip install accelerate2import torch
3from transformers import AutoTokenizer, AutoModelForCausalLM
45tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-mamba-7b-instruct")6model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-mamba-7b-instruct", device_map="auto", torch_dtype=torch.float16)78# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating9messages =[10{"role":"user","content":"How many helicopters can a human eat in one sitting?"},11]12input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)13input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")1415outputs = model.generate(input_ids, max_new_tokens=30)16print(tokenizer.decode(outputs[0]))
4-bit
Click to expand
python
1# pip install bitsandbytes accelerate2from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
34tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-mamba-7b-instruct")5model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-mamba-7b-instruct", device_map="auto", quantization_config=BitsAndBytesConfig(load_in_4bit=True))67# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating8messages =[9{"role":"user","content":"How many helicopters can a human eat in one sitting?"},10]11input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)12input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")1314outputs = model.generate(input_ids, max_new_tokens=30)15print(tokenizer.decode(outputs[0]))
Training Details
Training Data
Falcon-Mamba has been trained with ~ 5,500 GT mainly coming from Refined-Web, a large volume web-only dataset filtered and deduplicated.
Similar to the others Falcon suite models, Falcon-Mamba has been trained leveraging a multi-stage training strategy to increase the context-length from 2,048 to 8,192.
Moreover, inspired by the concept of Curriculum Learning, we carefully selected data mixtures throughout the training stages, considering both data diversity and complexity.
Note that at inference the context-length is not relevant as the Mamba architecture has no limit on long range dependency.
At the last training stage, small portion of high-quality curated data was used to further enhance performance.
Overall, the data sources included RefinedWeb-English, high quality technical data, code data and math data extracted from public sources.
In particular, we used samples coming from Fineweb-edu during our last training stage.
The data was tokenized with the Falcon-7B/11B tokenizer.
After pre-training, the model has been further fine-tuned on instruction data.
Training Procedure
Falcon-Mamba-7B was trained on 256 H100 80GB GPUs for the majority of the training, using a 3D parallelism strategy (TP=1, PP=1, DP=256) combined with ZeRO.
Training Hyperparameters
Hyperparameter
Value
Comment
Precision
bfloat16
Optimizer
AdamW
Max learning rate
6.4e-4
Following a WSD (warmup-stable-decay) learning rate schedule
Weight decay
1e-1
Batch size
2048
The model was trained AdamW optimizer, WSD (warmup-stable-decay) learning rate schedule, and a batch size rampup from \(b_{\mathrm{min}}=128\) to \(b_{\mathrm{max}}=2048\) during first 50 GT of training.
In the stable phase we used maximal learning rate \(\eta_{\mathrm{max}}=6.4 \times 10^{-4}\), and decayed it to the minimal value \(\eta_{\mathrm{min}}=\frac{\eta_{\mathrm{max}}}{256}\) with exponential schedule over 500 GT.
Also, we applied BatchScaling during the rampup — rescaling learning rate \(\eta\) so that the Adam noise temperature \(T_{\mathrm{noise}}\equiv\frac{\eta}{\sqrt{b}}\) is kept constant.
Speeds, Sizes, Times
The model training took roughly two months.
Evaluation
Benchmarks
We evaluate our model on all benchmarks of the new leaderboard's version using the lm-evaluation-harness package, and then normalize the evaluation results with HuggingFace score normalization.
model name
IFEval
BBH
MATH LvL5
GPQA
MUSR
MMLU-PRO
Average
Pure SSM models
FalconMamba-7B
33.36
19.88
3.63
8.05
10.86
14.47
15.04
TRI-ML/mamba-7b-rw*
22.46
6.71
0.45
1.12
5.51
1.69
6.25
Hybrid SSM-attention models
recurrentgemma-9b
30.76
14.80
4.83
4.70
6.60
17.88
13.20
Zyphra/Zamba-7B-v1*
24.06
21.12
3.32
3.03
7.74
16.02
12.55
Transformer models
Falcon2-11B
32.61
21.94
2.34
2.80
7.53
15.44
13.78
Meta-Llama-3-8B
14.55
24.50
3.25
7.38
6.24
24.55
13.41
Meta-Llama-3.1-8B
12.70
25.29
4.61
6.15
8.98
24.95
13.78
Mistral-7B-v0.1
23.86
22.02
2.49
5.59
10.68
22.36
14.50
Mistral-Nemo-Base-2407 (12B)
16.83
29.37
4.98
5.82
6.52
27.46
15.08
gemma-7B
26.59
21.12
6.42
4.92
10.98
21.64
15.28
Also, we evaluate our model on the benchmarks of the first leaderboard using lighteval.
model name
ARC
HellaSwag
MMLU
Winogrande
TruthfulQA
GSM8K
Average
Pure SSM models
FalconMamba-7B*
62.03
80.82
62.11
73.64
53.42
52.54
64.09
TRI-ML/mamba-7b-rw*
51.25
80.85
33.41
71.11
32.08
4.70
45.52
Hybrid SSM-attention models
recurrentgemma-9b**
52.00
80.40
60.50
73.60
38.60
42.60
57.95
Zyphra/Zamba-7B-v1*
56.14
82.23
58.11
79.87
52.88
30.78
60.00
Transformer models
Falcon2-11B
59.73
82.91
58.37
78.30
52.56
53.83
64.28
Meta-Llama-3-8B
60.24
82.23
66.70
78.45
42.93
45.19
62.62
Meta-Llama-3.1-8B
58.53
82.13
66.43
74.35
44.29
47.92
62.28
Mistral-7B-v0.1
59.98
83.31
64.16
78.37
42.15
37.83
60.97
gemma-7B
61.09
82.20
64.56
79.01
44.79
50.87
63.75
Mostly, we took evaluation results from both leaderboards. For the models marked by star we evaluated the tasks internally, while for the models marked by two stars the results were taken from paper or model card.
Throughput
This model can achieve comparable throughput and performance compared to other transformer based models that use optimized kernels such as Flash Attention 2. Make sure to install the optimized Mamba kernels with the following commands:
pip install "causal-conv1d>=1.4.0" mamba-ssm
Refer to our FalconMamba blogpost for more details about performance evaluation.
Technical Specifications
Model Architecture and Objective
Falcon-Mamba-7B is a causal decoder-only model trained on a causal language modeling task (i.e., predict the next token).
The model is based on the Mamba architecture (Gu et al., 2023).
Hyperparameter
Value
Comment
Layers
64
Number of layers
d_model
4096
Hidden dimension
d_state
16
The SSM state dimension
Vocabulary
65024
Vocabulary Size
Sequence length
8192
During the last training stages
Compute Infrastructure
Hardware
Falcon-Mamba-7B was trained on AWS SageMaker, using on average 256 H100 80GB GPUs in 32 p5 instances.
Software
Falcon-Mamba-7B was trained on an internal distributed training codebase, Gigatron. It uses a 3D parallelism approach combined with ZeRO, high-performance Triton kernels.
Citation
You can use the following bibtex citation:
@misc{zuo2024falconmambacompetitiveattentionfree,
title={Falcon Mamba: The First Competitive Attention-free 7B Language Model},
author={Jingwei Zuo and Maksim Velikanov and Dhia Eddine Rhaiem and Ilyas Chahed and Younes Belkada and Guillaume Kunsch and Hakim Hacid},
year={2024},
eprint={2410.05355},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2410.05355},
}
