Doron Adler
commited on
Commit
•
bffdb0a
1
Parent(s):
68eb283
Added inference examples using the .pt and .onnx models
Browse files- examples/example-onnx-infer.py +30 -0
- examples/example-pt-infer.py +22 -0
examples/example-onnx-infer.py
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
#Tested with the following Python package versions:
|
3 |
+
#optimum 1.2.3.dev0
|
4 |
+
#transformers 4.21.0.dev0
|
5 |
+
#tokenizers 0.11.6
|
6 |
+
|
7 |
+
from transformers import AutoTokenizer
|
8 |
+
from optimum.onnxruntime import ORTModelForCausalLM
|
9 |
+
from optimum.pipelines import pipeline
|
10 |
+
|
11 |
+
|
12 |
+
def main():
|
13 |
+
model_name="Norod78/distilgpt2-base-pretrained-he"
|
14 |
+
|
15 |
+
prompt_text = "שלום, קוראים לי"
|
16 |
+
generated_max_length = 192
|
17 |
+
|
18 |
+
print("Loading model...")
|
19 |
+
model = ORTModelForCausalLM.from_pretrained(model_name)
|
20 |
+
print('Loading Tokenizer...')
|
21 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
22 |
+
text_generator = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
|
23 |
+
|
24 |
+
print("Generating text...")
|
25 |
+
result = text_generator(prompt_text, num_return_sequences=1, batch_size=1, do_sample=True, top_k=40, top_p=0.92, temperature = 1, repetition_penalty=5.0, max_length = generated_max_length)
|
26 |
+
|
27 |
+
print("result = " + str(result))
|
28 |
+
|
29 |
+
if __name__ == '__main__':
|
30 |
+
main()
|
examples/example-pt-infer.py
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
3 |
+
|
4 |
+
def main():
|
5 |
+
model_name="Norod78/distilgpt2-base-pretrained-he"
|
6 |
+
|
7 |
+
prompt_text = "שלום, קוראים לי"
|
8 |
+
generated_max_length = 192
|
9 |
+
|
10 |
+
print("Loading model...")
|
11 |
+
model = AutoModelForCausalLM.from_pretrained(model_name)
|
12 |
+
print('Loading Tokenizer...')
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
14 |
+
text_generator = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
|
15 |
+
|
16 |
+
print("Generating text...")
|
17 |
+
result = text_generator(prompt_text, num_return_sequences=1, batch_size=1, do_sample=True, top_k=40, top_p=0.92, temperature = 1, repetition_penalty=5.0, max_length = generated_max_length)
|
18 |
+
|
19 |
+
print("result = " + str(result))
|
20 |
+
|
21 |
+
if __name__ == '__main__':
|
22 |
+
main()
|