Remove quantization_config entirely to avoid bitsandbytes dependency3

app.py

@@ -1,38 +1,13 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
-import accelerate
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
-print("Accelerate version:", accelerate.__version__)
+# Load the original model
+original_model = AutoModelForCausalLM.from_pretrained("unsloth/DeepSeek-R1-Distill-Llama-8B")
+original_tokenizer = AutoTokenizer.from_pretrained("unsloth/DeepSeek-R1-Distill-Llama-8B")
 
-# Load the original model with overridden configuration
-original_config = AutoConfig.from_pretrained(
-    "unsloth/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit"
-)
-# Remove quantization-related attributes from the config
-original_config._load_in_4bit = False
-original_config._load_in_8bit = False
-original_config.quant_method = None
-
-original_model = AutoModelForCausalLM.from_pretrained(
-    "unsloth/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit",
-    config=original_config  # Use the overridden configuration
-)
-original_tokenizer = AutoTokenizer.from_pretrained("unsloth/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit")
-
-# Load the fine-tuned model with overridden configuration
-fine_tuned_config = AutoConfig.from_pretrained(
-    "kas1/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit-John1"
-)
-# Remove quantization-related attributes from the config
-fine_tuned_config._load_in_4bit = False
-fine_tuned_config._load_in_8bit = False
-fine_tuned_config.quant_method = None
-
-fine_tuned_model = AutoModelForCausalLM.from_pretrained(
-    "kas1/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit-John1",
-    config=fine_tuned_config  # Use the overridden configuration
-)
-fine_tuned_tokenizer = AutoTokenizer.from_pretrained("kas1/DeepSeek-R1-Distill-Llama-8B-unsloth-bnb-4bit-John1")
+# Load the fine-tuned model
+fine_tuned_model = AutoModelForCausalLM.from_pretrained("kas1/DeepSeek-R1-Distill-Llama-8B-John1")
+fine_tuned_tokenizer = AutoTokenizer.from_pretrained("kas1/DeepSeek-R1-Distill-Llama-8B-John1")
 
 # Function to generate responses from both models
 def compare_models(prompt):

requirements.txt

@@ -1,4 +1,3 @@
 torch
 transformers
-gradio
-accelerate>=0.26.0
+gradio