Spaces:
Runtime error
Runtime error
jordigonzm
commited on
Commit
•
c86f71e
1
Parent(s):
9932057
Update app.py
Browse files
app.py
CHANGED
@@ -8,39 +8,35 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
8 |
|
9 |
model_name = "microsoft/Phi-3-mini-128k-instruct"
|
10 |
|
|
|
|
|
|
|
11 |
pipeline = transformers.pipeline(
|
12 |
"text-generation",
|
13 |
-
model=
|
14 |
-
|
15 |
device="cpu",
|
16 |
trust_remote_code=True
|
17 |
)
|
18 |
|
19 |
@spaces.GPU
|
20 |
def chat_function(message, history, system_prompt,max_new_tokens,temperature):
|
21 |
-
|
22 |
-
|
23 |
-
{"role": "
|
24 |
-
|
25 |
-
|
26 |
-
messages,
|
27 |
-
tokenize=False,
|
28 |
-
add_generation_prompt=True
|
29 |
-
)
|
30 |
-
terminators = [
|
31 |
-
pipeline.tokenizer.eos_token_id,
|
32 |
-
pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
|
33 |
-
]
|
34 |
temp = temperature + 0.1
|
35 |
outputs = pipeline(
|
36 |
prompt,
|
37 |
max_new_tokens=max_new_tokens,
|
38 |
-
eos_token_id=terminators,
|
39 |
do_sample=True,
|
40 |
temperature=temp,
|
41 |
top_p=0.9,
|
42 |
)
|
43 |
-
|
|
|
|
|
44 |
|
45 |
gr.ChatInterface(
|
46 |
chat_function,
|
|
|
8 |
|
9 |
model_name = "microsoft/Phi-3-mini-128k-instruct"
|
10 |
|
11 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
12 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16)
|
13 |
+
|
14 |
pipeline = transformers.pipeline(
|
15 |
"text-generation",
|
16 |
+
model=model,
|
17 |
+
tokenizer=tokenizer,
|
18 |
device="cpu",
|
19 |
trust_remote_code=True
|
20 |
)
|
21 |
|
22 |
@spaces.GPU
|
23 |
def chat_function(message, history, system_prompt,max_new_tokens,temperature):
|
24 |
+
prompt = system_prompt
|
25 |
+
for msg in history:
|
26 |
+
prompt += f'{msg["role"]}: {msg["content"]}\n'
|
27 |
+
prompt += f'user: {message}\n'
|
28 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
temp = temperature + 0.1
|
30 |
outputs = pipeline(
|
31 |
prompt,
|
32 |
max_new_tokens=max_new_tokens,
|
|
|
33 |
do_sample=True,
|
34 |
temperature=temp,
|
35 |
top_p=0.9,
|
36 |
)
|
37 |
+
generated_text = outputs[0]['generated_text']
|
38 |
+
new_text = generated_text[len(prompt):]
|
39 |
+
return new_text
|
40 |
|
41 |
gr.ChatInterface(
|
42 |
chat_function,
|