deepseek R1

app.py

@@ -1,28 +1,71 @@
 import streamlit as st
 import os
-from langchain_huggingface import HuggingFaceEndpoint
+from dotenv import load_dotenv
+from transformers import (
+    pipeline,
+    AutoConfig,
+    AutoTokenizer,
+    AutoModelForCausalLM
+)
+
+# Load environment variables from .env (if you’re using one)
+load_dotenv()
 
 st.set_page_config(page_title="Educational Chatbot")
 st.title("🎓 Educational Chatbot")
 
-user_input = st.text_input("Ask me anything:")
+@st.cache_resource(show_spinner=False)
+def load_model():
+    # 1. Load the remote config (with trust_remote_code)
+    config = AutoConfig.from_pretrained(
+        "deepseek-ai/DeepSeek-R1",
+        trust_remote_code=True
+    )
+    # 2. Remove unsupported fp8 quantization
+    if hasattr(config, "quantization_config"):
+        config.quantization_config = None
+
+    # 3. Load tokenizer and model with patched config
+    tokenizer = AutoTokenizer.from_pretrained(
+        "deepseek-ai/DeepSeek-R1",
+        trust_remote_code=True
+    )
+    model = AutoModelForCausalLM.from_pretrained(
+        "deepseek-ai/DeepSeek-R1",
+        trust_remote_code=True,
+        config=config
+    )
+
+    # 4. Build the text-generation pipeline
+    gen = pipeline(
+        "text-generation",
+        model=model,
+        tokenizer=tokenizer,
+        trust_remote_code=True,
+        device_map="auto"        # or remove for CPU-only
+    )
+    return gen
 
+# Load the model once
+generator = load_model()
+
+# Initialize chat history
 if "history" not in st.session_state:
     st.session_state.history = []
 
+# User input box
+user_input = st.text_input("Ask me anything:")
+
+# When user enters a question
 if user_input:
     try:
-        llm = HuggingFaceEndpoint(
-            repo_id="mistralai/Mistral-7B-Instruct-v0.1",
-            huggingfacehub_api_token=os.environ.get("HUGGINGFACEHUB_API_TOKEN"),
-            temperature=0.5,
-            max_new_tokens=100,
-        )
-        response = llm.invoke(user_input)
+        outputs = generator(user_input, return_full_text=False)
+        reply = outputs[0]["generated_text"].strip()
         st.session_state.history.append(("You", user_input))
-        st.session_state.history.append(("Bot", response))
+        st.session_state.history.append(("Bot", reply))
     except Exception as e:
-        st.error(f"⚠️ Error: {e}")
+        st.session_state.history.append(("Bot", f"⚠️ Error: {e}"))
 
+# Display chat history
 for sender, msg in reversed(st.session_state.history):
     st.markdown(f"**{sender}:** {msg}")

requirements.txt

@@ -1,6 +1,5 @@
-streamlit
-langchain
-langchain-community
-langchain-huggingface
-huggingface_hub
+streamlit>=1.20.0
 python-dotenv
+transformers>=4.40.0
+torch>=2.1.0
+accelerate