Possibly Incorrect Loss Computation on Chronos-Bolt?

[quang-truong]

I wonder if the following loss computation is intentional or typo, since it is quite different from that of Chronos-2? Specifically, shouldn't the first mean operation be over the dim -1?

chronos-forecasting/src/chronos/chronos_bolt.py

Lines 343 to 345 in f951d9a

	loss = loss.mean(dim=-2) # Mean over prediction horizon
	loss = loss.sum(dim=-1) # Sum over quantile levels
	loss = loss.mean() # Mean over batch

[abdulfatir]

See: #312