mirror of https://github.com/InternLM/InternLM
refactor
parent
ccdaf8ec45
commit
fd932cfc09
|
@ -132,7 +132,6 @@ runs_bak/
|
||||||
LLM_ALERT
|
LLM_ALERT
|
||||||
small_demo/
|
small_demo/
|
||||||
7b_llama_nopp/
|
7b_llama_nopp/
|
||||||
test/
|
|
||||||
|
|
||||||
# Pytorch
|
# Pytorch
|
||||||
*.pth
|
*.pth
|
||||||
|
|
|
@ -50,4 +50,4 @@ repos:
|
||||||
[
|
[
|
||||||
'--rcfile=.pylintrc',
|
'--rcfile=.pylintrc',
|
||||||
'--disable=C0114,C0415,W0212,W0235,W0238,W0621,C0103,R1735,C2801,E0402,C0412,W0719,R1728,W1514,W0718,W0105,W0707,C0209,W0703,W1203'
|
'--disable=C0114,C0415,W0212,W0235,W0238,W0621,C0103,R1735,C2801,E0402,C0412,W0719,R1728,W1514,W0718,W0105,W0707,C0209,W0703,W1203'
|
||||||
]
|
]
|
|
@ -186,11 +186,11 @@ def train(
|
||||||
# do forward and backward
|
# do forward and backward
|
||||||
timer("fwd-bwd").start()
|
timer("fwd-bwd").start()
|
||||||
|
|
||||||
_, _, loss, _ = trainer.execute_schedule(batch, forward_only=False, return_loss=True, return_output_label=False)
|
_, _, loss, moe_loss = trainer.execute_schedule(batch, forward_only=False, return_loss=True, return_output_label=False)
|
||||||
if gpc.is_rank_for_log():
|
if gpc.is_rank_for_log():
|
||||||
assert loss is not None and not math.isnan(loss.item())
|
assert loss is not None and not math.isnan(loss.item())
|
||||||
global cur_loss_list
|
global cur_loss_list
|
||||||
cur_loss_list.append(loss.item())
|
cur_loss_list.append(loss.item() - moe_loss.item())
|
||||||
timer("fwd-bwd").stop()
|
timer("fwd-bwd").stop()
|
||||||
|
|
||||||
# update parameters, and returns (success_update, grad_norm)
|
# update parameters, and returns (success_update, grad_norm)
|
||||||
|
|
Loading…
Reference in New Issue