update model card README.md

README.md

@@ -156,7 +156,7 @@ The following hyperparameters were used during training:
                                                           'top_k': 0,
                                                           'top_p': 0.9},
                                       'name': 'unconditional',
-                                      'num_samples': 2560,
+                                      'num_samples': 512,
                                       'prefix': '<|aligned|>'},
                                      {'generate_kwargs': {'bad_words_ids': [[50257],
                                                                             [50258],
@@ -169,7 +169,7 @@ The following hyperparameters were used during training:
                                                           'top_k': 0,
                                                           'top_p': 0.9},
                                       'name': 'challenging_rtp',
-                                      'num_samples': 1024,
+                                      'num_samples': 512,
                                       'prefix': '<|aligned|>',
                                       'prompt_before_control': True,
                                       'prompts_path': 'resources/challenging_rtp.jsonl'}],
@@ -177,7 +177,7 @@ The following hyperparameters were used during training:
  'kl_gpt3_callback': {'force_call_on': [21362],
                       'gpt3_kwargs': {'model_name': 'davinci'},
                       'max_tokens': 64,
-                      'num_samples': 2048,
+                      'num_samples': 256,
                       'prefix': '<|aligned|>',
                       'should_insert_prefix': True},
  'model': {'from_scratch': True,
@@ -212,4 +212,4 @@ The following hyperparameters were used during training:
               'weight_decay': 0.1}}
 
 # Wandb URL:
-https://wandb.ai/kejian/uncategorized/runs/3m4axm31
+https://wandb.ai/kejian/uncategorized/runs/1gr4oro2