only break long lines in unpacking

[t-vi]

No description provided.

[carmocca]

I don't think you updated the backward trace. Here's what I used to check:

from litgpt import GPT
import thunder
import torch

# Use only two layers to keep the traces shorter for the demonstration
model = GPT.from_name("Llama-2-7b-hf", n_layer=2).cuda()
model = thunder.jit(model)
x = torch.randint(model.max_seq_length, (2, 5), device="cuda")
y = model(x)  # forward, this may take a bit

forward_trace = thunder.last_traces(model)[-1]
print(forward_trace)
trace = thunder.last_backward_traces(model)[-1]
print(trace)

[t-vi]

Thank you @carmocca

[carmocca]

Amazing!

[lantiga]

Stamped!

[carmocca]

I didn't rerun my snippet after your last commits 😅. Opened #69