Upload 2 files

app.py

@@ -1,22 +1,14 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 
-# 配置 4-bit 量化以适配 16GB GPU
-quantization_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    bnb_4bit_compute_dtype=torch.float16,
-    bnb_4bit_quant_type="nf4",
-    bnb_4bit_use_double_quant=True
-)
-
-# 加载 DeepSeek-Coder-6.7B-Instruct 模型
-model_name = "deepseek-ai/deepseek-coder-6.7b-instruct"
-tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+# 加载轻量模型（CPU 兼容）
+tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/deepseek-coder-1.3b-base")
+model = AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-coder-1.3b-base")
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    quantization_config=quantization_config,
-    device_map="auto",
+    torch_dtype=torch.float16,  # 使用 FP16 减少内存
+    device_map="cpu",          # 强制使用 CPU
     trust_remote_code=True,
     low_cpu_mem_usage=True
 )
@@ -41,7 +33,7 @@ def respond(
 
     # 使用聊天模板格式化输入
     input_text = tokenizer.apply_chat_template(messages, tokenize=False)
-    inputs = tokenizer(input_text, return_tensors="pt").to("cuda")
+    inputs = tokenizer(input_text, return_tensors="pt").to("cpu")
 
     # 生成响应
     outputs = model.generate(

requirements.txt

@@ -1,5 +1,5 @@
-torch
-transformers
-accelerate
-gradio
-bitsandbytes  # Optional, for 4-bit quantization
+torch>=2.0.0
+transformers>=4.38.0
+accelerate>=0.26.0
+gradio>=4.0.0
+bitsandbytes>=0.43.3