Update handler.py
Browse files- handler.py +2 -6
handler.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
from transformers import AutoTokenizer, AutoModelForCausalLM
|
2 |
from transformers_stream_generator import init_stream_support
|
3 |
import re
|
4 |
init_stream_support()
|
@@ -22,16 +22,12 @@ Alice Gate: *Alice strides into the room with a smile, her eyes lighting up when
|
|
22 |
|
23 |
class EndpointHandler():
|
24 |
|
25 |
-
def __init__(self, path = "
|
26 |
-
path = "."
|
27 |
-
# quantization_config = BitsAndBytesConfig(llm_int8_enable_fp32_cpu_offload = True)
|
28 |
self.tokenizer = AutoTokenizer.from_pretrained(path)
|
29 |
self.model = AutoModelForCausalLM.from_pretrained(
|
30 |
path,
|
31 |
device_map = "auto",
|
32 |
load_in_8bit = True,
|
33 |
-
torch_dtype = "auto",
|
34 |
-
low_cpu_mem_usage = True
|
35 |
)
|
36 |
|
37 |
def __call__(self, data):
|
|
|
1 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
2 |
from transformers_stream_generator import init_stream_support
|
3 |
import re
|
4 |
init_stream_support()
|
|
|
22 |
|
23 |
class EndpointHandler():
|
24 |
|
25 |
+
def __init__(self, path = ""):
|
|
|
|
|
26 |
self.tokenizer = AutoTokenizer.from_pretrained(path)
|
27 |
self.model = AutoModelForCausalLM.from_pretrained(
|
28 |
path,
|
29 |
device_map = "auto",
|
30 |
load_in_8bit = True,
|
|
|
|
|
31 |
)
|
32 |
|
33 |
def __call__(self, data):
|