jrahn commited on
Commit
57bf368
1 Parent(s): 1ff4000

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -97,8 +97,8 @@ Masked Language Modeling objective with 15% masked token ratio.
97
 
98
  ### Preprocessing
99
 
100
- Tokenize `data["train"]["fen"]` with max-length padding to 200 tokens with default `distilbert-base-cased` tokenizer. Experiments with reduced max-length in tokenization show performance gains.
101
- Reducing vocab and sequence length in the model architecture shows gains in resource efficiency, but lacks in learning performance.
102
 
103
  ### Speeds, Sizes, Times
104
 
 
97
 
98
  ### Preprocessing
99
 
100
+ Tokenize `data["train"]["fen"]` with max-length padding to 200 tokens with default `distilbert-base-cased` tokenizer.
101
+ Experiments with reduced max-length in tokenization show performance gains.
102
 
103
  ### Speeds, Sizes, Times
104