ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

LLM: disable mmap by default for better performance (#8467)

Browse source
This commit is contained in:
Yishuo Wang 2023-07-11 09:26:26 +08:00 committed by GitHub
parent 52c6b057d6
commit db39d0a6b3
4 changed files with 6 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
python/llm/src/bigdl/llm/ggml/model/bloom/bloom.py
View file

@ -68,7 +68,7 @@ class Bloom(GenerationMixin):
f16_kv: bool = True, f16_kv: bool = True,
logits_all: bool = False, logits_all: bool = False,
vocab_only: bool = False, vocab_only: bool = False,
use_mmap: bool = True, use_mmap: bool = False,
use_mlock: bool = False, use_mlock: bool = False,
embedding: bool = False, embedding: bool = False,
n_threads: Optional[int] = 2, n_threads: Optional[int] = 2,
@ -127,7 +127,7 @@ class Bloom(GenerationMixin):
self.verbose = verbose self.verbose = verbose
# TODO: Some parameters are temporarily not supported # TODO: Some parameters are temporarily not supported
unsupported_arg = {'n_parts': -1, 'n_gpu_layers': 0, 'f16_kv': True, 'logits_all': False, unsupported_arg = {'n_parts': -1, 'n_gpu_layers': 0, 'f16_kv': True, 'logits_all': False,
'vocab_only': False, 'use_mmap': True, 'use_mlock': False, 'vocab_only': False, 'use_mmap': False, 'use_mlock': False,
'last_n_tokens_size': 64, 'lora_base': None, 'last_n_tokens_size': 64, 'lora_base': None,
'lora_path': None, 'verbose': True} 'lora_path': None, 'verbose': True}
for arg in unsupported_arg.keys(): for arg in unsupported_arg.keys():

2
python/llm/src/bigdl/llm/ggml/model/gptneox/gptneox.py
View file

@ -136,7 +136,7 @@ class Gptneox(GenerationMixin):
f16_kv: bool = True, f16_kv: bool = True,
logits_all: bool = False, logits_all: bool = False,
vocab_only: bool = False, vocab_only: bool = False,
use_mmap: bool = True, use_mmap: bool = False,
use_mlock: bool = False, use_mlock: bool = False,
embedding: bool = False, embedding: bool = False,
n_threads: Optional[int] = 2, n_threads: Optional[int] = 2,

2
python/llm/src/bigdl/llm/ggml/model/llama/llama.py
View file

@ -134,7 +134,7 @@ class Llama(GenerationMixin):
f16_kv: bool = True, f16_kv: bool = True,
logits_all: bool = False, logits_all: bool = False,
vocab_only: bool = False, vocab_only: bool = False,
use_mmap: bool = True, use_mmap: bool = False,
use_mlock: bool = False, use_mlock: bool = False,
embedding: bool = False, embedding: bool = False,
n_threads: Optional[int] = 2, n_threads: Optional[int] = 2,

4
python/llm/src/bigdl/llm/ggml/model/starcoder/starcoder.py
View file

@ -69,7 +69,7 @@ class Starcoder(GenerationMixin):
f16_kv: bool = True, f16_kv: bool = True,
logits_all: bool = False, logits_all: bool = False,
vocab_only: bool = False, vocab_only: bool = False,
use_mmap: bool = True, use_mmap: bool = False,
use_mlock: bool = False, use_mlock: bool = False,
embedding: bool = False, embedding: bool = False,
n_threads: Optional[int] = 2, n_threads: Optional[int] = 2,
@ -128,7 +128,7 @@ class Starcoder(GenerationMixin):
self.verbose = verbose self.verbose = verbose
# TODO: Some parameters are temporarily not supported # TODO: Some parameters are temporarily not supported
unsupported_arg = {'n_parts': -1, 'n_gpu_layers': 0, 'f16_kv': True, 'logits_all': False, unsupported_arg = {'n_parts': -1, 'n_gpu_layers': 0, 'f16_kv': True, 'logits_all': False,
'vocab_only': False, 'use_mmap': True, 'use_mlock': False, 'vocab_only': False, 'use_mmap': False, 'use_mlock': False,
'last_n_tokens_size': 64, 'lora_base': None, 'last_n_tokens_size': 64, 'lora_base': None,
'lora_path': None, 'verbose': True} 'lora_path': None, 'verbose': True}
for arg in unsupported_arg.keys(): for arg in unsupported_arg.keys():