This is a 'hacked' version of
tiiuae/Falcon3-10B-Base-1.58bit
where model weight scales have been injected into ternary model weights in order to make the model compatible with fine-tuning
The model has been trained following the training strategies from the recent
1-bit LLM HF blogpost
and
1-bit LLM paper
.
For more details about the training protocol of this model, please refer to the Falcon-3 technical report, section
Compression
.
Usage
Currently to use this model you can either rely on Hugging Face transformers library or
BitNet
library. You can also play with the model using the
falcon-1.58bit playground
(only for the 7B instruct version).
🤗 transformers
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "tiiuae/Falcon3-10B-Base-1.58bit"
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
).to("cuda")
# Perform text generation
BitNet
git clone https://github.com/microsoft/BitNet && cd BitNet
pip install -r requirements.txt
python setup_env.py --hf-repo tiiuae/Falcon3-10B-Base-1.58bit -q i2_s
python run_inference.py -m models/Falcon3-10B-1.58bit/ggml-model-i2_s.gguf -p "You are a helpful assistant" -cnv
Evaluation
We report in the following table our internal pipeline benchmarks:
Note evaluation results are normalized score from v2 leaderboard tasks - reported results of original models in the blogpost are raw scores
Benchmark
Llama3-8B-1.58-100B-tokens
Falcon3-10B-Base-1.58bit
IFEval
17.91
24.89
MUSR
4.87
4.6
GPQA
1.83
1.83
BBH
5.36
4.44
MMLU-PRO
2.78
1.36
MATH
0.26
0.48
Average
5.5
6.27
Citation
Coming soon ..
Runs of tiiuae Falcon3-10B-Base-1.58bit-prequantized on huggingface.co
175
Total runs
0
24-hour runs
4
3-day runs
43
7-day runs
47
30-day runs
More Information About Falcon3-10B-Base-1.58bit-prequantized huggingface.co Model
More Falcon3-10B-Base-1.58bit-prequantized license Visit here:
Falcon3-10B-Base-1.58bit-prequantized huggingface.co is an AI model on huggingface.co that provides Falcon3-10B-Base-1.58bit-prequantized's model effect (), which can be used instantly with this tiiuae Falcon3-10B-Base-1.58bit-prequantized model. huggingface.co supports a free trial of the Falcon3-10B-Base-1.58bit-prequantized model, and also provides paid use of the Falcon3-10B-Base-1.58bit-prequantized. Support call Falcon3-10B-Base-1.58bit-prequantized model through api, including Node.js, Python, http.
Falcon3-10B-Base-1.58bit-prequantized huggingface.co is an online trial and call api platform, which integrates Falcon3-10B-Base-1.58bit-prequantized's modeling effects, including api services, and provides a free online trial of Falcon3-10B-Base-1.58bit-prequantized, you can try Falcon3-10B-Base-1.58bit-prequantized online for free by clicking the link below.
tiiuae Falcon3-10B-Base-1.58bit-prequantized online free url in huggingface.co:
Falcon3-10B-Base-1.58bit-prequantized is an open source model from GitHub that offers a free installation service, and any user can find Falcon3-10B-Base-1.58bit-prequantized on GitHub to install. At the same time, huggingface.co provides the effect of Falcon3-10B-Base-1.58bit-prequantized install, users can directly use Falcon3-10B-Base-1.58bit-prequantized installed effect in huggingface.co for debugging and trial. It also supports api for free installation.
Falcon3-10B-Base-1.58bit-prequantized install url in huggingface.co: