Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Commit5b80748

Browse files
authored
update huggingface dependencies and calls (#1013)
1 parent280f017 commit5b80748

File tree

5 files changed

+32
-23
lines changed

5 files changed

+32
-23
lines changed

‎pgml-dashboard/content/blog/announcing-gptq-and-ggml-quantized-llm-support-for-huggingface-transformers.md‎

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -104,7 +104,10 @@ PostgresML will automatically use GPTQ or GGML when a HuggingFace model has one
104104
SELECTpgml.transform(
105105
task=>'{
106106
"task": "text-generation",
107-
"model": "mlabonne/gpt2-GPTQ-4bit"
107+
"model": "mlabonne/gpt2-GPTQ-4bit",
108+
"model_basename": "gptq_model-4bit-128g",
109+
"use_triton": true,
110+
"use_safetensors": true
108111
}'::JSONB,
109112
inputs=> ARRAY[
110113
'Once upon a time,'

‎pgml-extension/examples/transformers.sql‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -90,14 +90,15 @@ SELECT pgml.transform(
9090
'Dominic Cobb is the foremost practitioner of the artistic science of extraction, inserting oneself into a subject''s dreams to obtain hidden information without the subject knowing, a concept taught to him by his professor father-in-law, Dr. Stephen Miles. Dom''s associates are Miles'' former students, who Dom requires as he has given up being the dream architect for reasons he won''t disclose. Dom''s primary associate, Arthur, believes it has something to do with Dom''s deceased wife, Mal, who often figures prominently and violently in those dreams, or Dom''s want to "go home" (get back to his own reality, which includes two young children). Dom''s work is generally in corporate espionage. As the subjects don''t want the information to get into the wrong hands, the clients have zero tolerance for failure. Dom is also a wanted man, as many of his past subjects have learned what Dom has done to them. One of those subjects, Mr. Saito, offers Dom a job he can''t refuse: to take the concept one step further into inception, namely planting thoughts into the subject''s dreams without them knowing. Inception can fundamentally alter that person as a being. Saito''s target is Robert Michael Fischer, the heir to an energy business empire, which has the potential to rule the world if continued on the current trajectory. Beyond the complex logistics of the dream architecture of the case and some unknowns concerning Fischer, the biggest obstacles in success for the team become worrying about one aspect of inception which Cobb fails to disclose to the other team members prior to the job, and Cobb''s newest associate Ariadne''s belief that Cobb''s own subconscious, especially as it relates to Mal, may be taking over what happens in the dreams.'
9191
]
9292
);
93+
9394
SELECTpgml.transform(
9495
task=>'{"task": "text-classification",
9596
"model": "finiteautomata/bertweet-base-sentiment-analysis"
9697
}'::JSONB,
9798
inputs=> ARRAY[
98-
'I love how amazingly simple ML has become!',
99+
'I love how amazingly simple ML has become!',
99100
'I hate doing mundane and thankless tasks. ☹️'
100-
],
101+
]
101102
)AS positivity;
102103

103104
SELECTpgml.transform(
Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1 +1 @@
1-
xformers==0.0.20
1+
xformers==0.0.21

‎pgml-extension/requirements.txt‎

Lines changed: 15 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,16 +1,16 @@
1-
accelerate==0.21.0
2-
auto-gptq==0.3.0
3-
bitsandbytes==0.40.2
1+
accelerate==0.22.0
2+
auto-gptq==0.4.2
3+
bitsandbytes==0.41.1
44
catboost==1.2
5-
ctransformers==0.2.13
6-
datasets==2.13.1
7-
deepspeed==0.10.0
8-
huggingface-hub==0.16.4
5+
ctransformers==0.2.27
6+
datasets==2.14.5
7+
deepspeed==0.10.3
8+
huggingface-hub==0.17.1
99
InstructorEmbedding==1.0.1
10-
lightgbm==4.0.0
11-
orjson==3.9.2
12-
pandas==2.0.3
13-
rich==13.4.2
10+
lightgbm==4.1.0
11+
orjson==3.9.7
12+
pandas==2.1.0
13+
rich==13.5.2
1414
rouge==1.0.1
1515
sacrebleu==2.3.1
1616
sacremoses==0.0.53
@@ -21,9 +21,9 @@ tokenizers==0.13.3
2121
torch==2.0.1
2222
torchaudio==2.0.2
2323
torchvision==0.15.2
24-
tqdm==4.65.0
25-
transformers==4.31.0
26-
xgboost==1.7.6
27-
langchain==0.0.237
24+
tqdm==4.66.1
25+
transformers==4.33.1
26+
xgboost==2.0.0
27+
langchain==0.0.287
2828
einops==0.6.1
2929
pynvml==11.5.0

‎pgml-extension/src/bindings/transformers/transformers.py‎

Lines changed: 9 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -89,11 +89,12 @@ def ensure_device(kwargs):
8989

9090
classGPTQPipeline(object):
9191
def__init__(self,model_name,**task):
92-
importauto_gptq
92+
fromauto_gptqimportAutoGPTQForCausalLM,BaseQuantizeConfig
9393
fromhuggingface_hubimportsnapshot_download
9494
model_path=snapshot_download(model_name)
9595

96-
self.model=auto_gptq.AutoGPTQForCausalLM.from_quantized(model_path,**task)
96+
quantized_config=BaseQuantizeConfig.from_pretrained(model_path)
97+
self.model=AutoGPTQForCausalLM.from_quantized(model_path,quantized_config=quantized_config,**task)
9798
if"use_fast_tokenizer"intask:
9899
self.tokenizer=AutoTokenizer.from_pretrained(model_path,use_fast=task.pop("use_fast_tokenizer"))
99100
else:
@@ -192,9 +193,13 @@ def create_pipeline(task):
192193
ensure_device(task)
193194
convert_dtype(task)
194195
model_name=task.get("model",None)
195-
ifmodel_nameand"-ggml"inmodel_name:
196+
ifmodel_name:
197+
lower=model_name.lower()
198+
else:
199+
lower=None
200+
iflowerand"-ggml"inlower:
196201
pipe=GGMLPipeline(model_name,**task)
197-
elifmodel_nameand"-gptq"inmodel_name:
202+
eliflowerand"-gptq"inlower:
198203
pipe=GPTQPipeline(model_name,**task)
199204
else:
200205
try:

0 commit comments

Comments
 (0)

[8]ページ先頭

©2009-2025 Movatter.jp