tokenization mismatch

#25
by ehartford - opened

I'm using fastchat to finetune, and I get errors like this:

WARNING: tokenization mismatch: 381 vs. 382. (ignored)
WARNING: tokenization mismatch: 411 vs. 412. (ignored)
WARNING: tokenization mismatch: 250 vs. 251. (ignored)
WARNING: tokenization mismatch: 117 vs. 119. (ignored)
WARNING: tokenization mismatch: 508 vs. 510. (ignored)
WARNING: tokenization mismatch: 511 vs. 512. (ignored)
WARNING: tokenization mismatch: 713 vs. 714. (ignored)
WARNING: tokenization mismatch: 253 vs. 255. (ignored)
WARNING: tokenization mismatch: 461 vs. 463. (ignored)
WARNING: tokenization mismatch: 233 vs. 234. (ignored)
WARNING: tokenization mismatch: 235 vs. 236. (ignored)
WARNING: tokenization mismatch: 2281 vs. 2282. (ignored)
WARNING: tokenization mismatch: 276 vs. 277. (ignored)
WARNING: tokenization mismatch: 907 vs. 908. (ignored)
WARNING: tokenization mismatch: 171 vs. 172. (ignored)
WARNING: tokenization mismatch: 360 vs. 361. (ignored)

Any idea?

Sign up or log in to comment