Update app.py
Browse files
app.py
CHANGED
|
@@ -5,7 +5,12 @@ import gradio as gr
|
|
| 5 |
import spaces
|
| 6 |
import torch
|
| 7 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
|
|
|
| 8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
MAX_MAX_NEW_TOKENS = 2048
|
| 10 |
DEFAULT_MAX_NEW_TOKENS = 1024
|
| 11 |
total_count=0
|
|
@@ -15,7 +20,7 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
|
|
| 15 |
DESCRIPTION="""CODE"""
|
| 16 |
|
| 17 |
model_id = "deepseek-ai/deepseek-coder-7b-instruct-v1.5"
|
| 18 |
-
model = AutoModelForCausalLM.from_pretrained(model_id,device_map="auto",
|
| 19 |
tokenizer=AutoTokenizer.from_pretrained(model_id)
|
| 20 |
tokenizer.use_defaul_system_prompt=True
|
| 21 |
|
|
|
|
| 5 |
import spaces
|
| 6 |
import torch
|
| 7 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
| 8 |
+
from transformers import BitsAndBytesConfig
|
| 9 |
|
| 10 |
+
nf4_config = BitsAndBytesConfig(
|
| 11 |
+
load_in_4bit=True,
|
| 12 |
+
bnb_4bit_quant_type="nf4",
|
| 13 |
+
)
|
| 14 |
MAX_MAX_NEW_TOKENS = 2048
|
| 15 |
DEFAULT_MAX_NEW_TOKENS = 1024
|
| 16 |
total_count=0
|
|
|
|
| 20 |
DESCRIPTION="""CODE"""
|
| 21 |
|
| 22 |
model_id = "deepseek-ai/deepseek-coder-7b-instruct-v1.5"
|
| 23 |
+
model = AutoModelForCausalLM.from_pretrained(model_id,device_map="auto",quantization_config=nf4_config)
|
| 24 |
tokenizer=AutoTokenizer.from_pretrained(model_id)
|
| 25 |
tokenizer.use_defaul_system_prompt=True
|
| 26 |
|