Update app.py

app.py

@@ -6,10 +6,10 @@ import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 
 # tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm2-6b-int4", trust_remote_code=True)
-# model = AutoModel.from_pretrained("THUDM/chatglm2-6b-int4", trust_remote_code=True)
-tokenizer = AutoTokenizer.from_pretrained("openchat/openchat_v2_w")
-model = AutoModelForCausalLM.from_pretrained("openchat/openchat_v2_w", load_in_8bit_fp32_cpu_offload=True, load_in_8bit=True)
-model.float()
+model = AutoModel.from_pretrained("THUDM/chatglm2-6b-int4", trust_remote_code=True)
+# tokenizer = AutoTokenizer.from_pretrained("openchat/openchat_v2_w")
+# model = AutoModelForCausalLM.from_pretrained("openchat/openchat_v2_w", load_in_8bit_fp32_cpu_offload=True, load_in_8bit=True)
+model.half()
 model = model.eval()
 
 model_path = model.config._dict['model_name_or_path']