[ Speculative decoding ] Support different tokenizers for draft and main models #7232
Job | Run time |
---|---|
18m 31s | |
31m 21s | |
17m 11s | |
33m 6s | |
20m 28s | |
12m 45s | |
20m 17s | |
36m 21s | |
9m 50s | |
15m 3s | |
28m 53s | |
22m 26s | |
15m 30s | |
26m 24s | |
12m 1s | |
7m 58s | |
14m 7s | |
7m 15s | |
12m 27s | |
29m 53s | |
17m 19s | |
1s | |
6h 49m 7s |