(fix): fix gspo config in doc

Author: breaddaerb

docs_roll/docs/English/UserGuide/algorithms/GSPO.md

@@ -27,7 +27,7 @@ response_length: 4096
 ppo_epochs: 1
 use_kl_loss: true
 kl_loss_coef: 0.001
-loss_agg_mode: "seq-mean-token-sum"
+loss_agg_mode: "seq-mean-token-mean"
 
 # advantage
 whiten_advantages: true

docs_roll/docs/简体中文/使用指南/algorithms/GSPO.md

@@ -27,7 +27,7 @@ response_length: 4096
 ppo_epochs: 1
 use_kl_loss: true
 kl_loss_coef: 0.001
-loss_agg_mode: "seq-mean-token-sum"
+loss_agg_mode: "seq-mean-token-mean"
 
 # advantage
 whiten_advantages: true

examples/docs_examples/example_gspo.yaml

@@ -53,7 +53,7 @@ response_length: 4096
 ppo_epochs: 1
 use_kl_loss: true
 kl_loss_coef: 0.001
-loss_agg_mode: "seq-mean-token-sum"
+loss_agg_mode: "seq-mean-token-mean"
 
 # advantage
 whiten_advantages: true