Fix the logic for perplexity evaluation (Not enough kv_cache capacity to run generation. Please use a larger sequence_length or a shorter prompt
)
#5381
Job | Run time |
---|---|
15m 10s | |
6m 18s | |
15m 16s | |
21m 51s | |
17m 2s | |
3m 47s | |
21m 10s | |
4m 15s | |
23m 16s | |
2h 8m 5s |