feat: set tokenizer

README.md

@@ -1,20 +1,3 @@
----
-title: SFT
-emoji: 🚀
-colorFrom: red
-colorTo: red
-sdk: docker
-app_port: 8501
-tags:
-- streamlit
-pinned: false
-short_description: Streamlit template space
-license: mit
----
+# CONOZ SFT Proj.
 
-# Welcome to Streamlit!
-
-Edit `/src/streamlit_app.py` to customize this app to your heart's desire. :heart:
-
-If you have any questions, checkout our [documentation](https://docs.streamlit.io) and [community
-forums](https://discuss.streamlit.io).
+meta-llama/Llama-3.1-8B -> choco-conoz/TwinLlama-3.1-8B

pyproject.toml

@@ -1,22 +0,0 @@
-[project]
-name = "src"
-version = "0.1.0"
-description = ""
-authors =  [
-    { name = "yongkyucho", email = "choco@conoz.net" },
-]
-license = "MIT"
-readme = "README.md"
-
-[build-system]
-requires = ["poetry-core>=2.0.0,<3.0.0"]
-build-backend = "poetry.core.masonry.api"
-
-[tool.poetry.dependencies]
-python = "~3.11"
-
-torch = "2.7.0"
-sentence-transformers = "^3.0.0"
-
-streamlit = "^1.46.1"
-# unsloth = "^2025.6.8"

src/streamlit_app.py

@@ -1,6 +1,6 @@
 import streamlit as st
 import torch
-from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, pipeline
 
 # model_id = "sentence-transformers/all-MiniLM-L6-v2"
 # model_id = "sentence-transformers/xlm-r-base-en-ko-nli-ststb"
@@ -38,14 +38,17 @@ def main():
 
     if st.button("Send"):
         user_prompt = alpaca_template.format(query, "")
-        print('user_prompt', user_prompt)
-        prompt = processor.tokenizer.apply_chat_template(
+        # print('user_prompt', user_prompt)
+        prompt = tokenizer.apply_chat_template(
             user_prompt, tokenize=False, add_generation_prompt=True)
         # prompt = user_prompt
         # outputs = processor(prompt)
-
-        outputs = processor(prompt, max_new_tokens=4096, eos_token_id=terminators, do_sample=True,
-                            temperature=0.6, top_p=0.9
+        outputs = processor(prompt,
+                            max_new_tokens=4096,
+                            eos_token_id=terminators,
+                            do_sample=True,
+                            temperature=0.6,
+                            top_p=0.9
                             )
         response = outputs[0]["generated_text"][len(prompt):]
         st.write(response)