Commit Graph

11 Commits

Author SHA1 Message Date
Nicolas Patry b965c5deee
Keeping the vocab to 99. 2022-02-23 11:44:13 +01:00
Nicolas Patry af496b2775
Update the tokenizer for 4.17. 2022-02-23 11:20:43 +01:00
Nicolas Patry 8cdd434766
Adding tf weights. 2022-02-17 14:50:44 +01:00
Nicolas Patry 76a74407e7
Removing all merges. 2022-02-16 18:17:38 +01:00
Nicolas Patry 12ceb59e69
Going the other way and simplifying the tokenizer instead. 2022-02-16 18:16:09 +01:00
Nicolas Patry 541d169ca9
Revert "Making the vocab_size match the tokenizer vocab_size."
This reverts commit 4be554b688.
2022-02-16 18:15:43 +01:00
Nicolas Patry 4be554b688
Making the vocab_size match the tokenizer vocab_size. 2022-02-16 18:14:27 +01:00
Nicolas Patry 10da33b5d6
crop_size is after resize ?? 2022-02-16 18:09:42 +01:00
Nicolas Patry 8de11cad3b
Adding pipeline_tag. 2022-02-16 18:05:41 +01:00
Nicolas Patry d05fe019ab
Making processor + model match. 2022-02-16 18:04:29 +01:00
system 809a93d113 initial commit 2022-02-16 17:03:38 +00:00