Toke(n)icer v0.0.4

What's Changed

⚡ Now tokenicer instance dynamically inherits the native tokenizer.__class__ of tokenizer passed in or loaded via our Tokenicer.load() api.
⚡ CI now tests tokenizers from 64 models

• fix mpt pad token bug by @CL-ModelCloud in #24
• fix model_config bugs by @CL-ModelCloud in #25
• test code clean up by @CL-ModelCloud in #26
• Inherits PretrainedTokenizer by @Qubitium in #28
• loop & test all models by @CSY-ModelCloud in #30

Full Changelog: v0.0.2...v0.0.4

Toke(n)icer v0.0.3

What's Changed

Now tokenicer instance dynamically inherits the native tokenizer.__class__ of tokenizer passed in or loaded via our Tokenicer.load() api.

• fix mpt pad token bug by @CL-ModelCloud in #24
• fix model_config bugs by @CL-ModelCloud in #25
• test code clean up by @CL-ModelCloud in #26
• Inherits PretrainedTokenizer by @Qubitium in #28

Full Changelog: v0.0.2...v0.0.3

Toke(n)icer v0.0.2

What's Changed

⚡ Auto-fix models not setting padding_token
⚡ Auto-Fix models released with wrong padding_token: many models incorrectly use eos_token as pad_token which leads to subtle and hidden errors in post-training and inference when batching is used which is almost always.
⚡ Compatible with all HF Transformers recognized tokenizers

• Auto fix pad token by @CL-ModelCloud in #5
• Forward to Tokenizer by @CL-ModelCloud in #6
• read requirements.txt in setup.py by @CSY-ModelCloud in #7
• [CI] add tokenicer forward test by @CL-ModelCloud in #10
• add unit tests by @CSY-ModelCloud in #11
• refractor by @Qubitium in #8
• add deepseek_v3 map by @CL-ModelCloud in #15

New Contributors

• @CSY-ModelCloud made their first contribution in #1
• @Qubitium made their first contribution in #3
• @CL-ModelCloud made their first contribution in #5

Full Changelog: https://github.com/ModelCloud/Tokenicer/commits/v0.0.2