Add gen_entities function
Browse files
app.py
CHANGED
@@ -11,16 +11,20 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
11 |
load_in_8bit=True,
|
12 |
device_map="auto",
|
13 |
revision="half",
|
14 |
-
# low_cpu_mem_usage=True
|
15 |
)
|
16 |
tokenizer = AutoTokenizer.from_pretrained(peft_model_id)
|
17 |
# Load the Lora model
|
18 |
model = PeftModel.from_pretrained(model, peft_model_id)
|
19 |
|
20 |
|
21 |
-
def
|
22 |
-
|
|
|
|
|
|
|
23 |
|
|
|
24 |
|
25 |
-
|
|
|
26 |
iface.launch()
|
|
|
11 |
load_in_8bit=True,
|
12 |
device_map="auto",
|
13 |
revision="half",
|
|
|
14 |
)
|
15 |
tokenizer = AutoTokenizer.from_pretrained(peft_model_id)
|
16 |
# Load the Lora model
|
17 |
model = PeftModel.from_pretrained(model, peft_model_id)
|
18 |
|
19 |
|
20 |
+
def gen_entities(text):
|
21 |
+
text = f"<SP> text: {text}\n\n entities: "
|
22 |
+
batch = tokenizer(text, return_tensors="pt")
|
23 |
+
with torch.cuda.amp.autocast():
|
24 |
+
output_tokens = model.generate(**batch, max_new_tokens=256, eos_token_id=50258)
|
25 |
|
26 |
+
return tokenizer.decode(output_tokens[0], skip_special_tokens=False)
|
27 |
|
28 |
+
|
29 |
+
iface = gr.Interface(fn=gen_entities, inputs="text", outputs="text")
|
30 |
iface.launch()
|