Hi @joanrod , Thanks for opensouring star vector!
I am trying to reimplement the exp results in your paper. However, I cannot run the eval code in validator since the code is incomplete.
e.g., when running eval with hf backend:
python starvector/validation/validate.py \
config=configs/generation/hf/starvector-1b/im2svg.yaml \
dataset.dataset_name=starvector/svg-stack
It raises error:
AttributeError: 'StarVectorHFSVGValidator' object has no attribute 'dataloader'. Did you mean: 'get_dataloader'?
When I add get_dataloader which is missing in your code:
it raises another error:
AttributeError: 'StarVectorHFSVGValidator' object has no attribute 'processor'
When running eval with vllm backend:
python starvector/validation/validate.py \
config=configs/generation/vllm/starvector-1b/im2svg.yaml \
dataset.dataset_name=starvector/svg-stack
it raises error:
[rank0]: ValueError: StarVectorForCausalLM has no vLLM implementation and the Transformers implementation is not compatible with vLLM.
Would you please look at the code and give me some advice about the fix? Thanks!
Hi @joanrod , Thanks for opensouring star vector!
I am trying to reimplement the exp results in your paper. However, I cannot run the eval code in
validatorsince the code is incomplete.e.g., when running eval with hf backend:
It raises error:
When I add
get_dataloaderwhich is missing in your code:it raises another error:
When running eval with vllm backend:
it raises error:
Would you please look at the code and give me some advice about the fix? Thanks!