Portx commited on
Commit
85ff963
·
verified ·
1 Parent(s): 8442a03

Update handler.py

Browse files
Files changed (1) hide show
  1. handler.py +3 -5
handler.py CHANGED
@@ -6,18 +6,16 @@ import os
6
 
7
  #run("pip install flash-attn --no-build-isolation", shell=True, check=True)
8
  run("pip install -U bitsandbytes", shell=True, check=True)
9
- def restart_script():
10
- python = sys.executable
11
- os.execl(python, python, *sys.argv)
12
 
13
- restart_script()
14
 
15
  from transformers import AutoModelForVision2Seq, AutoProcessor, BitsAndBytesConfig
16
 
17
  model_id = "ibm-granite/granite-vision-3.2-2b"
18
 
19
  bnb_config = BitsAndBytesConfig(
20
- load_in_8bit=True,
21
  llm_int8_skip_modules=["vision_tower", "lm_head"],
22
  llm_int8_enable_fp32_cpu_offload=True
23
  )
 
6
 
7
  #run("pip install flash-attn --no-build-isolation", shell=True, check=True)
8
  run("pip install -U bitsandbytes", shell=True, check=True)
9
+ run("pip install --upgrade pip", shell=True, check=True)
10
+
 
11
 
 
12
 
13
  from transformers import AutoModelForVision2Seq, AutoProcessor, BitsAndBytesConfig
14
 
15
  model_id = "ibm-granite/granite-vision-3.2-2b"
16
 
17
  bnb_config = BitsAndBytesConfig(
18
+ load_in_4bit=True,
19
  llm_int8_skip_modules=["vision_tower", "lm_head"],
20
  llm_int8_enable_fp32_cpu_offload=True
21
  )