wandb / utils.py
tcapelle's picture
put more context
c725b52
import random, time
import requests
import wandb
word_site = "https://www.mit.edu/~ecprice/wordlist.10000"
response = requests.get(word_site)
WORDS = [w.decode("UTF-8") for w in response.content.splitlines()]
def train(name, project="st", entity=None, epochs=10, bar=None):
run = wandb.init(
# Set the project where this run will be logged
name=name,
project=project,
entity=entity,
# Track hyperparameters and run metadata
config={
"learning_rate": 0.02,
"architecture": "CNN",
"dataset": "CIFAR-100",
"epochs": epochs,
})
# This simple block simulates a training loop logging metrics
offset = random.random() / 5
for epoch in range(1, epochs+1):
acc = 1 - 2 ** -epoch - random.random() / epoch - offset
loss = 2 ** -epoch + random.random() / epoch + offset
# 2️⃣ Log metrics from your script to W&B
wandb.log({"acc": acc, "loss": loss})
time.sleep(0.1)
bar.progress(epoch/epochs)
# Mark the run as finished
wandb.finish()