EchoStreet
/

mpt-7b

Text Generation

StreamingDatasets

Inference Endpoints

text-generation-inference

Model card Files Files and versions Community

rlanner-echocap commited on Jul 28, 2023

Commit

f9a0290

•

1 Parent(s): 09d8478

Update handler.py

Removing load in 8-bit flag

Files changed (1) hide show

handler.py +1 -1

handler.py CHANGED Viewed

@@ -11,7 +11,7 @@ class EndpointHandler:
     def __init__(self, path=""):
         # load the model
         tokenizer = AutoTokenizer.from_pretrained(path)
-        model = AutoModelForCausalLM.from_pretrained(path, load_in_8bit=True, device_map="auto", torch_dtype=dtype, trust_remote_code=True)
         # create inference pipeline
         self.pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer, device='cuda:0')

     def __init__(self, path=""):
         # load the model
         tokenizer = AutoTokenizer.from_pretrained(path)
+        model = AutoModelForCausalLM.from_pretrained(path, device_map="auto", torch_dtype=dtype, trust_remote_code=True)
         # create inference pipeline
         self.pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer, device='cuda:0')