Skip to content

Commit

Permalink
chore: use "BatchLongest"
Browse files Browse the repository at this point in the history
  • Loading branch information
Anush008 committed Jan 31, 2024
1 parent 25993ce commit 770a12e
Showing 1 changed file with 3 additions and 3 deletions.
6 changes: 3 additions & 3 deletions fastembed.go
Original file line number Diff line number Diff line change
Expand Up @@ -362,13 +362,13 @@ func loadTokenizer(modelPath string, maxLength int) (*tokenizer.Tokenizer, error
Stride: 0,
})

paddingStrategy := tokenizer.NewPaddingStrategy(tokenizer.WithFixed(maxLength))

paddingParams := tokenizer.PaddingParams{
Strategy: *paddingStrategy,
// Strategy defaults to "BatchLongest"
Strategy: *tokenizer.NewPaddingStrategy(),
Direction: tokenizer.Right,
PadId: int(config["pad_token_id"].(float64)),
PadToken: tokenizerConfig["pad_token"].(string),
PadTypeId: 0,
}
tknzer.WithPadding(&paddingParams)

Expand Down

0 comments on commit 770a12e

Please sign in to comment.