List your company

Language Modeling on OpenWebText GPT-2 (test)

17.94PerplexityAdamCPR (Kappa-IP)
17.902818.153918.40518.6561Nov 15, 2023

Evaluation results

MethodLinks
AdamCPR (Kappa-IP)Number of random seeds=3, Model=GPT2s2023.1117.94—
AdamCPR (Kappa-WS)Warm start steps (s)=5k (1x), Number of random seeds=3, Model=GPT2s2023.1118.020.0258
AdamCPR (Kappa-WS)Warm start steps (s)=10k (2x), Number of random seeds=3, Model=GPT2s2023.1118.030.0178
AdamWWeight decay coefficient (gamma)=1e-2, Number of random seeds=3, Model=GPT2s2023.1118.230.0113
AdamCPR (Kappa-WS)Warm start steps (s)=20k (4x), Number of random seeds=3, Model=GPT2s2023.1118.240.032
Adam AdadecayWeight decay coefficient (gamma)=1e-2, Number of random seeds=1, Model=GPT2s2023.1118.24—
Adam AdadecayWeight decay coefficient (gamma)=1e-3, Number of random seeds=1, Model=GPT2s2023.1118.42—
Adam AWDWeight decay coefficient (gamma)=1e-3, Number of random seeds=1, Model=GPT2s2023.1118.42—
AdamWWeight decay coefficient (gamma)=1e-3, Number of random seeds=3, Model=GPT2s2023.1118.450.0039
Adam AWDWeight decay coefficient (gamma)=1e-2, Number of random seeds=1, Model=GPT2s2023.1118.47—
Adam AWDWeight decay coefficient (gamma)=1e-1, Number of random seeds=1, Model=GPT2s2023.1118.49—
AdamWWeight decay coefficient (gamma)=1e-1, Number of random seeds=3, Model=GPT2s2023.1118.860.0169
Adam AdadecayWeight decay coefficient (gamma)=1e-1, Number of random seeds=1, Model=GPT2s2023.1118.87—