github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-05 19:13:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

[Coati] Train DPO using PP (#6054)

Browse Source 
* update dpo

* remove unsupport plugin

* update msg

* update dpo

* remove unsupport plugin

* update msg

* update template

* update dataset

* add pp for dpo

* update dpo

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* add dpo fn

* update dpo

* update dpo

* update dpo

* update dpo

* minor update

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* update loss

* update help

* polish code

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This commit is contained in:
Tong Li
2024-10-11 19:32:00 +08:00
committed by GitHub
parent dc2cdaf3e8
commit 4c8e85ee0d
8 changed files with 529 additions and 236 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
applications/ColossalChat/coati/models/loss.py
View File

@@ -153,10 +153,11 @@ class DpoLoss(nn.Module):
else:
# If no reference model is provided
ref_logratios = 0.0
pi_logratios = logprob_actor_chosen.sum(-1) - logprob_actor_reject.sum(-1)
logits = pi_logratios - ref_logratios - self.gamma / self.beta
losses = -torch.nn.functional.logsigmoid(self.beta * logits)
loss = losses.mean()
# Calculate rewards for logging
if logprob_ref_chosen is not None:
chosen_rewards = self.beta * (logprob_actor_chosen.sum(-1) - logprob_ref_chosen.sum(-1)).detach()
@@ -167,7 +168,7 @@ class DpoLoss(nn.Module):
else:
rejected_rewards = self.beta * logprob_actor_reject.sum(-1).detach()
return losses, chosen_rewards, rejected_rewards
return loss, chosen_rewards, rejected_rewards
class LogSigLoss(nn.Module):