The created model is not reachable

#1
by NikolayL - opened

I've successfully created a model using the Space: https://huggingface.co/NikolayL/TinyLlama-1.1B-Chat-v1.0-openvino-int4
But I can't launch it.
I've tried both, a pipeline and direct usage. They all failing on downloading the model with error:

OSError: NikolayL/TinyLlama-1.1B-Chat-v1.0-openvino-int4 does not appear to have a file named pytorch_model.bin, model.safetensors, tf_model.h5, model.ckpt or flax_model.msgpack.

image.png

I've entered my email and token with huggingface-cli before it. Anything else should be done in order to launch the model?

Interestingly, the tokenizer is reachable.

tokenizer = AutoTokenizer.from_pretrained(
    "NikolayL/TinyLlama-1.1B-Chat-v1.0-openvino-int4"
)

But when it comes to the model, the errors occurs:

model = AutoModelForCausalLM.from_pretrained(
    "NikolayL/TinyLlama-1.1B-Chat-v1.0-openvino-int4"
)
OpenVINO Toolkit org

Hi @NikolayL ,

You can load your model using optimum as described on your model card :

from optimum.intel import OVModelForCausalLM

model_id = "NikolayL/TinyLlama-1.1B-Chat-v1.0-openvino-int4"
model = OVModelForCausalLM.from_pretrained(model_id)

to install optimum you can do the following :

pip install optimum[openvino]

Hi @echarlaix ,

Thanks for the answer. Yes, model card's recipe works. But the problem is that "Use this model button" provides a different recipe with Auto classes, rather than with Optimum wrappers.
This might be confusing for users.
image.png

OpenVINO Toolkit org

Yes the code snippet from "use this model" should use optimum instead, thanks for letting me know, will take a look !

echarlaix changed discussion status to closed

Sign up or log in to comment