DeepPavlov

DeepPavlov copied to clipboard

DeepPavlov version (you can look it up by running pip show deeppavlov): Name: deeppavlov Version: 0.10.0

Python version: Python 3.7.7

Operating system (ubuntu linux, windows, ...): Arch linux

Issue: When i try to run kbqa model I got an error: https://pastebin.com/20d1YfJr What's going on and how to overcome it?

Content or a name of a configuration file:

kbqa_cq

Command that led to error:

python3 deeppavlov/deep.py interact kbqa_cq

Error (including full traceback):

https://pastebin.com/20d1YfJr

The component RelRanker requires GPU because the model uses CudnnGRU, this bug will be fixed in the next release (optional usage of CudnnCompatibleGRU or CudnnGRU depending on whether the user uses GPU or not).

How much of GPU memory I need to run RelRanker? I have installed tf-gpu and it seems I got memory allocation error. I have only GeForce 940MX with 2GB of memory.

Errors like that:

tensorflow.python.framework.errors_impl.ResourceExhaustedError: OOM when allocating tensor with shape[512,768] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc
	 [[{{node Optimizer/Optimizer/bert/embeddings/position_embeddings/AdamWeightDecayOptimizer/Assign}}]]
Hint: If you want to see a list of allocated tensors when OOM happens, add report_tensor_allocations_upon_oom to RunOptions for current allocation info.

And lot of that's:

2020-07-08 22:09:27.752969: I tensorflow/core/common_runtime/bfc_allocator.cc:917] 1 Chunks of size 169494528 totalling 161.64MiB
2020-07-08 22:09:27.752978: I tensorflow/core/common_runtime/bfc_allocator.cc:921] Sum Total of in-use chunks: 1.71GiB
2020-07-08 22:09:27.752987: I tensorflow/core/common_runtime/bfc_allocator.cc:923] total_region_allocated_bytes_: 1833959424 memory_limit_: 1833959424 available bytes: 0 curr_region_allocation_bytes_: 2147483648
2020-07-08 22:09:27.753001: I tensorflow/core/common_runtime/bfc_allocator.cc:929] Stats: 
Limit:                  1833959424
InUse:                  1833959424
MaxInUse:               1833959424
NumAllocs:                    1547
MaxAllocSize:            169494528

@exelents, sorry for the late response. We tested kbqa_cq_en config from the DeepPavlov latest version 1.2.0 and didn’t encounter the described problem. If you are still having difficulties with this, please let us know.