engine/scripts/testgen_vision.py at main

mirror of https://github.com/zenlm/engine.git synced 2026-10-30 15:00:05 +00:00

Files

T

* Update image_seq_len

* Update the examples

* Format

2024-08-23 10:43:14 -04:00

 import transformers
 processor = transformers.AutoProcessor.from_pretrained(
     "microsoft/Phi-3.5-vision-instruct", trust_remote_code=True
 )
 res = processor.tokenizer.apply_chat_template(
     [
         {
             "role": "system",
             "content": [{"type": "text", "text": "You are a helpful assistant"}],
         },
         {
             "role": "user",
             "content": [
                 {"type": "image"},
                 {
                     "type": "text",
                     "text": "Hello, please describe the above.",
                 },
             ],
         },
         {"role": "assistant", "content": [{"type": "text", "text": "Hi there"}]},
         {
             "role": "user",
             "content": [
                 {"type": "image"},
                 {"type": "text", "text": "This is me, who are you"},
             ],
         },
         {
             "role": "assistant",
             "content": [{"type": "text", "text": "   I am an assistant   "}],
         },
         {
             "role": "user",
             "content": [
                 {"type": "image"},
                 {"type": "text", "text": "Another question, what is this?"},
             ],
         },
     ],
     add_generation_prompt=True,
     tokenize=False,
 )
 print(res.replace("\n", "\\n"))