fr
AlphAI - Unlock the power of accelerated compute
pip install alphai
and start up your remote GPU servers, prototype AI and LLMs, benchmark code, and analyze profiling data.
Python Client for your remote servers
Use AlphAI to start up GPU servers with a Jupyter Lab IDE with any number of runtime environments that support PyTorch, Hugging Face, and more.
from alphai import AlphAI
api_key = ***************
aai = AlphAI(
api_key=os.environ.get("ALPHAI_API_KEY")
)
# Start remote Jupyter Lab servers
aai.start_server()
# Upload to your server's file system
aai.upload("./main.py")
# Start python kernel and run code remotely
code = "print('Hello world!')"
aai.run_code(code)
GPU Profiling
Analyze and profile tensor traces on your deep learning runs in your GPU servers.
from alphai import AlphAI
import torch
import math
aai = AlphAI()
model = torch.nn.Sequential(
torch.nn.Linear(3, 1),
torch.nn.Flatten(0, 1)
).to("cuda")
x = torch.linspace(-math.pi, math.pi, 2000)
p = torch.tensor([1, 2, 3])
xx = x.unsqueeze(-1).pow(p).to("cuda")
aai.start()
model(xx)
aai.stop()
Pythonic Benchmarking and Timing
Benchmark and time your code and functions.
from alphai import AlphAI
prompt = "Hello there!"
aai.start_timer()
aai.generate(prompt)
aai.stop_timer()
aai.benchmark(aai.generate, prompt, num_iter = 5)