fix the logit overflow caused by pad_token https://github.com/asahi417/lmppl/issues/5

experiments/{summary_validation.py → get_qualitative_table.py}

@@ -2,6 +2,9 @@ import json
 import pandas as pd
 from datasets import load_dataset
 
+pd.set_option('display.max_rows', None)
+pd.set_option('display.max_columns', None)
+
 data_valid = load_dataset("cardiffnlp/relentless", split="validation")
 lc_valid = pd.read_csv("results_validation/lm_lc/lm.csv", index_col=0)
 qa_valid = pd.read_csv("results_validation/lm_qa/lm.csv", index_col=0)
@@ -27,6 +30,11 @@ p = 30
 table = []
 for prompt in ['qa', 'lc']:
     for i in target.keys():
+        if i in ['flan-t5-xxl', 'flan-ul2'] and prompt == 'lc':
+            continue
+        if i in ['opt-13b', 'davinci'] and prompt == 'qa':
+            continue
+
         for d in data_test:
             with open(f"results/lm_{prompt}/{i}/ppl.{d['relation_type'].replace(' ', '_').replace('/', '__')}.jsonl") as f:
                 negative_ppl = sorted([json.loads(x)['perplexity'] * -1 for x in f.read().split("\n") if len(x) > 0], reverse=True)
@@ -48,4 +56,4 @@ for prompt in ['qa', 'lc']:
             })
 
 table = pd.DataFrame(table)
-# table.to_csv("results_validation/summary_validation.csv")
+table.to_csv("results_validation/qualitative.csv", index=False)

experiments/results_validation/qualitative.csv

@@ -0,0 +1,13 @@
+,,top,bottom
+"Flan-T5	extsubscript{XXL}",Ally,Darth Vader:Obi-Wan Kenobi,Maximus Decimus Meridius:Juba
+,Inf,,Vape:cigarette
+Flan-UL2,Ally,,Maximus Decimus Meridius:Juba
+,Inf,Rihanna:Stevie Wonder,
+,Know,Qualcomm:smartphones,
+"OPT	extsubscript{13B}",Rival,,"Wickes:ScrewFix, Russia:US"
+,Ally,"Darth Vader:Obi-Wan Kenobi, Schindler's List:Baz Luhrmann",Microsoft:OpenAI
+,Inf,Rihanna:Stevie Wonder,"UK:Winston Churchill, Vape:cigarette"
+,Sim,Domino's:YO! Sushi,
+"GPT-3	extsubscript{davinci}",Rival,,Wickes:ScrewFix
+,Ally,Darth Vader:Obi-Wan Kenobi,
+,Inf,Rihanna:Stevie Wonder,"Vape:cigarette, Liz Truss:Thatcher"