push

app.py

@@ -8,7 +8,7 @@ from accelerate import init_empty_weights, load_checkpoint_and_dispatch
 
 # nstruct_pipeline_3b = pipeline(model="fnlp/moss-moon-003-sft-int4", torch_dtype=torch.float, trust_remote_code=True,
 #                                device_map="auto")
-model_path = "fnlp/moss-moon-003-sft-int4"
+model_path = "fnlp/moss-moon-003-sft-int8"
 
 config = MossConfig.from_pretrained(model_path)
 tokenizer = MossTokenizer.from_pretrained(model_path)