iicd/ddpo-pytorch
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix stat tracking bug

Browse Source
This commit is contained in:
Kevin Black
2023-06-26 22:25:43 -07:00
parent 5c16a90ceb
commit 1ce0994c8a
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ddpo_pytorch/stat_tracking.py
View File

@@ -9,6 +9,8 @@ class PerPromptStatTracker:
self.stats = {} self.stats = {}
def update(self, prompts, rewards): def update(self, prompts, rewards):
prompts = np.array(prompts)
rewards = np.array(rewards)
unique = np.unique(prompts) unique = np.unique(prompts)
advantages = np.empty_like(rewards) advantages = np.empty_like(rewards)
for prompt in unique: for prompt in unique: