Upload 2 files

app.py

@@ -15,7 +15,6 @@ torch.set_float32_matmul_precision("high")
 HF_TOKEN = os.getenv("HF_TOKEN", None)
 #REPO_ID = "deepseek-ai/DeepSeek-R1-Distill-Llama-8B"
 REPO_ID = "nicoboss/DeepSeek-R1-Distill-Qwen-32B-Uncensored"
-#REPO_ID = "Qwen/QwQ-32B"
 
 DESCRIPTION = f'''
 <div>
@@ -71,8 +70,7 @@ def chat(message: str,
         messages.append({"role": "system", "content": sys_prompt})
         messages.append({"role": "user", "content": message})
 
-        #input_tensors = tokenizer.apply_chat_template(history + messages, add_generation_prompt=True, return_dict=True, return_tensors="pt").to(model.device)
-        input_tensors = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_dict=True, return_tensors="pt").to(model.device)
+        input_tensors = tokenizer.apply_chat_template([{"role": x["role"], "content": x["content"]} for x in history] + messages, add_generation_prompt=True, return_dict=True, add_special_tokens=False, return_tensors="pt").to(model.device)
 
         input_ids = input_tensors["input_ids"]
         attention_mask = input_tensors["attention_mask"]

requirements.txt

@@ -6,6 +6,4 @@ git+https://github.com/huggingface/transformers
 numpy<2
 sentencepiece
 triton
-optimum
-optimum-quanto
 bitsandbytes