Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -16,23 +16,9 @@ def respond(
|
|
16 |
temperature,
|
17 |
top_p,
|
18 |
):
|
19 |
-
quantization_config = BitsAndBytesConfig(
|
20 |
-
bnb_4bit_compute_dtype="float16",
|
21 |
-
bnb_4bit_quant_storage="uint8",
|
22 |
-
bnb_4bit_quant_type="nf4",
|
23 |
-
bnb_4bit_use_double_quant=True,
|
24 |
-
llm_int8_enable_fp32_cpu_offload=False,
|
25 |
-
llm_int8_has_fp16_weight=False,
|
26 |
-
llm_int8_skip_modules=None,
|
27 |
-
llm_int8_threshold=6.0,
|
28 |
-
load_in_4bit=True,
|
29 |
-
load_in_8bit=False,
|
30 |
-
quant_method="bitsandbytes"
|
31 |
-
)
|
32 |
-
|
33 |
# Load model and tokenizer
|
34 |
model = AutoModelForCausalLM.from_pretrained("Rorical/0-roleplay", return_dict=True, trust_remote_code=True)
|
35 |
-
tokenizer = AutoTokenizer.from_pretrained("Rorical/0-roleplay", trust_remote_code=True
|
36 |
tokenizer.chat_template = "{% for message in messages %}{{'<|im_start|>' + ((message['role'] + ':\n') if message['role'] != '' else '') + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>ζι:\n' }}{% endif %}" # Be careful that this model used custom chat template.
|
37 |
|
38 |
# Construct the messages for the chat
|
|
|
16 |
temperature,
|
17 |
top_p,
|
18 |
):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
# Load model and tokenizer
|
20 |
model = AutoModelForCausalLM.from_pretrained("Rorical/0-roleplay", return_dict=True, trust_remote_code=True)
|
21 |
+
tokenizer = AutoTokenizer.from_pretrained("Rorical/0-roleplay", trust_remote_code=True)
|
22 |
tokenizer.chat_template = "{% for message in messages %}{{'<|im_start|>' + ((message['role'] + ':\n') if message['role'] != '' else '') + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>ζι:\n' }}{% endif %}" # Be careful that this model used custom chat template.
|
23 |
|
24 |
# Construct the messages for the chat
|