Slower generation with multi-batch size.
#26 opened 1 day ago
by
Satandon1999
why only small model use MuP ?
#24 opened 3 days ago
by
xdseunghyun
tokenizer gives unexpected results
#23 opened 7 days ago
by
jpiabrantes
Does it support system?
#22 opened 8 days ago
by
lucasjin
Upload triton_flash_blocksparse_attn.py
#21 opened 9 days ago
by
barcelosallan
"TypeError: phi3 isn't supported yet ". Could not quantize the phi3 model using the AWQ quantization method.
3
#20 opened 16 days ago
by
rabinghimire737
Multi-GPU case device mismatch while finetuning.
3
#19 opened 23 days ago
by
Satandon1999
CheckpointError in `triton_flash_blocksparse_attn.py` while finetuning
1
#18 opened 25 days ago
by
FremyCompany
Fix for FlashAttention RuntimeError & Triton Multi GPU fix.
1
#17 opened about 1 month ago
by
Satandon1999
Out of resource: shared memory
1
#16 opened about 1 month ago
by
iszhaoxin
RuntimeError: FlashAttention only support fp16 and bf16 data type
1
#15 opened about 1 month ago
by
Satandon1999
Adding sample_finetune.py to small model
1
#9 opened about 2 months ago
by
hackint0sh
The attention mask and the pad token id were not set
#8 opened about 2 months ago
by
hamidpalangi