failed to load Tensor shape expected by framework [] doesn't match model configuration shape [-1]
See original GitHub issueI am trying to load the model which is text and doesn’t have specific input shape this is my config file:
name: "model"
platform: "tensorflow_savedmodel"
max_batch_size: 16
input [
{
name: "label_ids"
data_type: TYPE_INT64
dims: []
},
{
name: "input_ids"
data_type: TYPE_INT64
dims: [128]
},
{
name: "input_mask"
data_type: TYPE_FP32
format: FORMAT_NHWC
dims: [128]
},
{
name: "segment_ids"
data_type: TYPE_FP32
format: FORMAT_NHWC
dims: [128]
}
]
output [
{
name: "output"
data_type: TYPE_FP32
dims: [ 1300 ]
}
]
instance_group [
{
count: 1
gpus: [0]
}
]
dynamic_batching {
preferred_batch_size: [ 16 ]
max_queue_delay_microseconds: 100
}
I know that labels_id doesn’t have specific shape, but when I give empty list I get this error:
E1003 18:35:18.944193 1 server.cc:137] model input must specify ‘dims’ for model
And when I try with [-1] I get this error:
failed to load ‘mode’ version 1: Invalid argument: unable to load model ‘model’, tensor ‘label_ids’ shape expected by framework [] doesn’t match model configuration shape [-1]
Any help?
My command:
**sudo nvidia-docker run --rm --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 -p8000:8000 -p8001:8001 -p8002:8002 -v/home/serving/teste/models_trt:/models nvcr.io/nvidia/tensorrtserver:19.06-py3 trtserver --model-store=/models --strict-model-config=false**
Issue Analytics
- State:
- Created 4 years ago
- Comments:6 (3 by maintainers)
Top GitHub Comments
@yoonjaecho This is an old closed issue. Can you file a new issue with your question and delete your comment from this issue.
@deadeyegoodwin thanks! reshaping worked.
Closing it!