generate error
#43
by
fz147258
- opened
Environment: GPU RTX 3090 * 2; python 3.10
I will Dolly_ v2_ 12b is encapsulated as an API, and errors will be reported when concurrent requests are made: mat1 and mat2 shapes cannot be multiplexed (1293x24 and 46x5120)
Not sure, maybe your input is too long?
Not sure, maybe your input is too long?
thanks，
but input hello error
The 3090 has 24gb?
Does it work if you just run this directly without your service?
How about just loading from HF?
Looks pretty ok otherwise
srowen
changed discussion status to
closed