Final Rescue: Full vocab restoration with Unigram scores and Metaspace fix a76cdb7 verified suchirsalhan commited on 6 days ago
Final Fix: Correct Metaspace mapping and Unigram scores 90447ca verified suchirsalhan commited on 6 days ago
Fix: Final Metaspace decoding using exact vocab marker cdd44c2 verified suchirsalhan commited on 6 days ago
Fix: Robust SPM-to-HF conversion with merges and byte-fallback 141ef67 verified suchirsalhan commited on 6 days ago
Fix: Final Metaspace decoding using exact vocab marker 422a108 verified suchirsalhan commited on 6 days ago
Fix: Reverted to native SentencePiece handling (removed ByteLevel mismatch) 22f93a2 verified suchirsalhan commited on 6 days ago
Fix: Applied ByteLevel pre-tokenization and decoding for proper spacing/hex handling cb9966c verified suchirsalhan commited on 6 days ago
Fix: Clean base tokenizer using universal metaspace decoding 14f9844 verified suchirsalhan commited on 6 days ago
Fix: Universal Metaspace decoding without keyword conflicts 08df509 verified suchirsalhan commited on 6 days ago
Fix: Manual vocab extraction from spm.model pieces 61176b3 verified suchirsalhan commited on 6 days ago
Fix: Environment issues resolved, full Fast Tokenizer uploaded. 9605799 verified suchirsalhan commited on 6 days ago