correcting the code for inference

.gitignore

@@ -0,0 +1 @@
+*.ipynb

README.md

@@ -78,16 +78,19 @@ To use one of the models, load the base `Qwen3-4B` model and then apply the adap
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
 from peft import PeftModel
+import os
 
-# Configuration
 base_model_name = "Qwen/Qwen3-4B"
 adapter_repo_id = "AxelDlv00/ToxiFrench" 
+target_adapter = "SOAP-DWL-DPO" 
 
-# Choose the target adapter
-# ["Standard-SFT", "SOAP-SFT", "SOAP-Oversampled", "SOAP-DWL", "SOAP-DWL-DPO"]
-target_adapter = "SOAP-DWL" 
+tokenizer = AutoTokenizer.from_pretrained(base_model_name, trust_remote_code=True)
+if tokenizer.pad_token is None:
+    tokenizer.pad_token = tokenizer.eos_token
+
+tokens = ["<think>", "</think>"]
+tokenizer.add_special_tokens({"additional_special_tokens": tokens})
 
-# Quantization
 bnb_config = BitsAndBytesConfig(
     load_in_4bit=True,
     bnb_4bit_use_double_quant=True,
@@ -95,23 +98,27 @@ bnb_config = BitsAndBytesConfig(
     bnb_4bit_compute_dtype=torch.bfloat16
 )
 
-# Load Tokenizer & Model
-tokenizer = AutoTokenizer.from_pretrained(base_model_name, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
     base_model_name,
     quantization_config=bnb_config,
-    trust_remote_code=True
+    trust_remote_code=True,
+    device_map="auto"
 )
 
-# Load the Adapter
+tokenizer_vocab_size = len(tokenizer)
+model_embedding_size = model.get_input_embeddings().weight.size(0)
+
+if model_embedding_size != tokenizer_vocab_size:
+    print(f"Syncing vocab: {model_embedding_size} -> {tokenizer_vocab_size}")
+    model.resize_token_embeddings(tokenizer_vocab_size)
+
 model = PeftModel.from_pretrained(model, adapter_repo_id, subfolder=target_adapter)
 model.eval()
 
-# Inference
 text = "Je ne supporte plus ton comportement, tu es vraiment un idiot !"
 prompt = f"Message:\n{text}\n\nAnalyse:\n"
-
 inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+
 with torch.no_grad():
     outputs = model.generate(
         **inputs,