2024-11-21 12:41:00
机器之心原创
扣子 Coze OpenAPI 不仅具备低延时、定制化、随时打断和音色克隆等优势,还整合了扣子强大的智能体生态。
© THE END
转载请联系本公众号获得授权
投稿或寻求报道:[email protected]
2024-11-21 12:41:00
2024-11-21 12:41:00
机器之心报道
编辑:泽南、小舟
租用 H100 的钱只需 233 美元。
pip install -r requirements.txt
pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu124 —upgrade # install torch 2.6.0
python data/cached_fineweb10B.py 10 # downloads only the first 1.0B training tokens to save time
./run.sh
sudo apt-get update
sudo apt-get install vim tmux python3-pip python-is-python3 -y
git clone https://github.com/KellerJordan/modded-nanogpt.git
cd modded-nanogpt
tmux
pip install numpy==1.23.5 huggingface-hub tqdm
pip install --upgrade torch &
python data/cached_fineweb10B.py 18
sudo docker build -t modded-nanogpt .
sudo docker run -it --rm --gpus all -v $(pwd):/modded-nanogpt modded-nanogpt python data/cached_fineweb10B.py 18
sudo docker run -it --rm --gpus all -v $(pwd):/modded-nanogpt modded-nanogpt sh run.sh
@torch.compile
def zeroth_power_via_newtonschulz5 (G, steps=5, eps=1e-7):
assert len (G.shape) == 2
b, c = (3.4445, -4.7750, 2.0315)
X = G.bfloat16 () / (G.norm () + eps)
if G.size (0) > G.size (1):
X = X.T
for _ in range (steps):
A = X @ X.T
B = b * A + c * A @ A
X = a * X + B @ X
if G.size (0) > G.size (1):
X = X.T
return X.to (G.dtype)
2024-11-21 12:41:00
机器之心报道
编辑:杜伟、陈陈
谷歌「Alpha」家族又壮大了,这次瞄准了量子计算领域。
2024-11-21 12:41:00
AIxiv专栏是机器之心发布学术、技术内容的栏目。过去数年,机器之心AIxiv专栏接收报道了2000多篇内容,覆盖全球各大高校与企业的顶级实验室,有效促进了学术交流与传播。如果您有优秀的工作想要分享,欢迎投稿或者联系报道。投稿邮箱:[email protected];[email protected]
论文地址: https://arxiv.org/pdf/2410.20418
代码仓库: https://github.com/xiaoniu-578fa6bff964d005/AcceleratedUnbiasedWatermark