text-generation-inference/server/text_generation_server/models
Nicolas Patry 1da07e85aa
feat(server): Add Non flash MPT. (#514)
# What does this PR do?


This adds a non flash version of MPT.
Flash is harder because we need to create a bias ready cuda kernel of
flash attention.

Fixes
https://github.com/huggingface/text-generation-inference/issues/361
Fixes
https://github.com/huggingface/text-generation-inference/issues/491
Fixes
https://github.com/huggingface/text-generation-inference/issues/290
2023-07-03 13:01:46 +02:00
..
custom_modeling feat(server): Add Non flash MPT. (#514) 2023-07-03 13:01:46 +02:00
__init__.py feat(server): Add Non flash MPT. (#514) 2023-07-03 13:01:46 +02:00
bloom.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
causal_lm.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
flash_causal_lm.py feat(server): add paged attention to flash models (#516) 2023-06-30 19:09:59 +02:00
flash_llama.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
flash_neox.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
flash_rw.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
flash_santacoder.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
galactica.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
gpt_neox.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
model.py feat(server): add paged attention to flash models (#516) 2023-06-30 19:09:59 +02:00
mpt.py feat(server): Add Non flash MPT. (#514) 2023-07-03 13:01:46 +02:00
opt.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
rw.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
santacoder.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
seq2seq_lm.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
t5.py feat: Add the option to force another dtype than f16. (#513) 2023-06-30 20:30:09 +02:00
types.py feat(server): support vectorized warpers in flash causal lm (#317) 2023-05-26 12:30:27 +02:00