Code for Keras plays catch blog post
python qlearn.py
- Generate figures
from transformers import AutoTokenizer, T5ForConditionalGeneration | |
# Model Init | |
n_gpu = 8 | |
tokenizer = AutoTokenizer.from_pretrained("google/flan-ul2") | |
model = T5ForConditionalGeneration.from_pretrained("google/flan-ul2") | |
heads_per_gpu = len(model.encoder.block) // n_gpu | |
device_map = { | |
gpu: list( | |
range( |
#!/bin/bash | |
# Invoke like this: | |
# ./watch.sh my command here | |
# And it will run 'my command here' once, and then when it detects changes. | |
# TODO: Don't just search in the last second. Search for updates since the last | |
# completed build. Otherwise for big directories, midway through your search | |
# you've already taken 1s and you will miss updates. |
Code for Keras plays catch blog post
python qlearn.py