Pszemraj's workspace
Runs
5
Name
5 visualized
Tags
loss
step
Notes
learning_rate
batch_size
wizard of wikipedia
1.50245
10454
-
0.0000042428
32
No EOS token
1.40829
2014
no eos token appended at the end
0.000042216
32
Chatbot
ai-msgbot
gpt-peter
1.48527
4139
the first training run of the GPT-peter dataset V5 on GPT-J 8bit
1.2385e-8
32
Chatbot
ai-msgbot
dailydialogue
1.49044
2723
single epoch
1.8825e-8
32
Chatbot
ai-msgbot
wizard of wikipedia
1.52689
1001
1000k steps or 5% of train
0.000049028
32
1-5
of 5