Spaces:
Sleeping
Sleeping
will.k
commited on
Commit
·
dc822cd
1
Parent(s):
556f311
app.py
CHANGED
@@ -16,18 +16,21 @@ peft_config = PeftConfig.from_pretrained("pseudolab/K23_MiniMed")
|
|
16 |
peft_model = MistralForCausalLM.from_pretrained("pseudolab/K23_MiniMed", trust_remote_code=True)
|
17 |
peft_model = PeftModel.from_pretrained(peft_model, "pseudolab/K23_MiniMed")
|
18 |
|
|
|
|
|
19 |
# Prepare the context
|
20 |
def prepare_context(data):
|
21 |
# Format the data as a string
|
22 |
data_str = data.to_string(index=False, header=False)
|
23 |
|
24 |
# Tokenize the data
|
25 |
-
input_ids = tokenizer.encode(data_str, return_tensors="pt")
|
26 |
|
27 |
# Truncate the input if it's too long for the model
|
28 |
-
max_length = tokenizer.model_max_length
|
29 |
-
if input_ids.shape[1] > max_length:
|
30 |
-
|
|
|
31 |
|
32 |
return input_ids
|
33 |
|
@@ -37,7 +40,8 @@ def fn(uploaded_file) -> str:
|
|
37 |
|
38 |
# Generate text based on the context
|
39 |
context = prepare_context(data)
|
40 |
-
generated_text = pipeline('text-generation', model=peft_model)(context)[0]['generated_text']
|
|
|
41 |
ret += generated_text
|
42 |
|
43 |
# Internally prompt the model to data analyze the EHR patient data
|
@@ -48,13 +52,14 @@ def fn(uploaded_file) -> str:
|
|
48 |
input_ids = tokenizer.encode(prompt, return_tensors="pt")
|
49 |
|
50 |
# Generate text based on the prompt
|
51 |
-
generated_text = pipeline('text-generation', model=peft_model)(input_ids=input_ids)[0]['generated_text']
|
|
|
52 |
ret += generated_text
|
53 |
|
54 |
return ret
|
55 |
|
56 |
|
57 |
-
demo = gr.Interface(fn=fn, inputs="file", outputs="text")
|
58 |
|
59 |
|
60 |
if __name__ == "__main__":
|
|
|
16 |
peft_model = MistralForCausalLM.from_pretrained("pseudolab/K23_MiniMed", trust_remote_code=True)
|
17 |
peft_model = PeftModel.from_pretrained(peft_model, "pseudolab/K23_MiniMed")
|
18 |
|
19 |
+
text_generator = pipeline('text-generation', model=peft_model, tokenizer=tokenizer)
|
20 |
+
|
21 |
# Prepare the context
|
22 |
def prepare_context(data):
|
23 |
# Format the data as a string
|
24 |
data_str = data.to_string(index=False, header=False)
|
25 |
|
26 |
# Tokenize the data
|
27 |
+
# input_ids = tokenizer.encode(data_str, return_tensors="pt")
|
28 |
|
29 |
# Truncate the input if it's too long for the model
|
30 |
+
# max_length = tokenizer.model_max_length
|
31 |
+
# if input_ids.shape[1] > max_length:
|
32 |
+
# input_ids = input_ids[:, :max_length]
|
33 |
+
input_ids = data_str
|
34 |
|
35 |
return input_ids
|
36 |
|
|
|
40 |
|
41 |
# Generate text based on the context
|
42 |
context = prepare_context(data)
|
43 |
+
# generated_text = pipeline('text-generation', model=peft_model, tokenizer=tokenizer)(context)[0]['generated_text']
|
44 |
+
generated_text = text_generator(context)[0]['generated_text']
|
45 |
ret += generated_text
|
46 |
|
47 |
# Internally prompt the model to data analyze the EHR patient data
|
|
|
52 |
input_ids = tokenizer.encode(prompt, return_tensors="pt")
|
53 |
|
54 |
# Generate text based on the prompt
|
55 |
+
# generated_text = pipeline('text-generation', model=peft_model, tokenizer=tokenizer)(input_ids=input_ids)[0]['generated_text']
|
56 |
+
generated_text = text_generator(prompt)[0]['generated_text']
|
57 |
ret += generated_text
|
58 |
|
59 |
return ret
|
60 |
|
61 |
|
62 |
+
demo = gr.Interface(fn=fn, inputs="file", outputs="text", theme="pseudolab/huggingface-korea-theme")
|
63 |
|
64 |
|
65 |
if __name__ == "__main__":
|