github
/
InternLM
mirror of https://github.com/InternLM/InternLM
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

refactor code

pull/448/head
Qu Wenwen 2023-10-27 15:34:48 +08:00
parent 739a308c82
commit 4df2c47472
1 changed files with 11 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
internlm/solver/optimizer/utils.py
View File

@ -252,16 +252,6 @@ def reduce_grads(gradients, parameters, fine_grained=False):
return parallel_grads return parallel_grads
def reduce_moe_norm(total_norm):
pg = gpc.get_group(ParallelMode.EXPERT)
scaled_norm = total_norm * 1.0 / float(gpc.get_world_size(ParallelMode.DATA))
scaled_norm_tensor = torch.tensor(scaled_norm, device=get_current_device(), dtype=torch.float)
dist.all_reduce(scaled_norm_tensor, group=pg)
total_norm = scaled_norm_tensor.item()
return total_norm
def compute_norm( def compute_norm(
gradients, gradients,
parameters, parameters,
@ -344,7 +334,11 @@ def compute_norm(
# Need to allreduce(avg) the norms across different ranks because moe params will not be synced during allreduce # Need to allreduce(avg) the norms across different ranks because moe params will not be synced during allreduce
# model and zero have been reduced!!! # model and zero have been reduced!!!
if zero_mode == ParallelMode.EXPERT_DATA: if zero_mode == ParallelMode.EXPERT_DATA:
total_norm = reduce_moe_norm(total_norm) pg = gpc.get_group(ParallelMode.EXPERT)
scaled_norm = total_norm * 1.0 / float(gpc.get_world_size(ParallelMode.DATA))
scaled_norm_tensor = torch.tensor(scaled_norm, device=get_current_device(), dtype=torch.float)
dist.all_reduce(scaled_norm_tensor, group=pg)
total_norm = scaled_norm_tensor.item()
# Scale. # Scale.
if total_norm == float("inf") or total_norm == -float("inf"): if total_norm == float("inf") or total_norm == -float("inf"):
@ -433,7 +427,12 @@ def compute_param_norm(
# moe # moe
if zero_mode == ParallelMode.EXPERT_DATA: if zero_mode == ParallelMode.EXPERT_DATA:
total_param_norms = reduce_moe_norm(total_param_norms) pg = gpc.get_group(ParallelMode.EXPERT)
scaled_param_norm = torch.cuda.FloatTensor(list(total_param_norms.values()), device=get_current_device())
scaled_param_norm = scaled_param_norm / float(gpc.get_world_size(ParallelMode.EXPERT))
dist.all_reduce(scaled_param_norm, group=pg)
for i, param_name in enumerate(total_param_norms.keys()):
total_param_norms[param_name] = scaled_param_norm[i].item()
# scale # scale
for param_name, param_norm in total_param_norms.items(): for param_name, param_norm in total_param_norms.items():