search

xiaowei-hu / pysc2-agents

This is a simple implementation of DeepMind's PySC2 RL agents.
https://zhuanlan.zhihu.com/p/29246185?group_id=890682069733232640
271 stars 77 forks source link

Tensor shape not matching #10

Open rwill128 opened 6 years ago

rwill128 commented 6 years ago

Hi,

I'm getting this error when I try to run the basic pre-trained agent available for download.

The call stack has my python version, my GPU, etc.

I'm using the newest GIT version of pysc2, which I think means my pysc2 and s2clientprotocol version are 1.2 instead of 1.1.

C:\Users\Rick\Desktop\pysc2-agents>python -m main --map=MoveToBeacon --training=false W0203 14:01:37.877619 1860 tf_logging.py:118] From C:\Users\Rick\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\ops\clip_ops.py:110: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version. Instructions for updating: keep_dims is deprecated, use keepdims instead 2018-02-03 14:01:38.338341: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\platform\cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2 2018-02-03 14:01:38.600618: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1105] Found device 0 with properties: name: GeForce GTX 1070 major: 6 minor: 1 memoryClockRate(GHz): 1.683 pciBusID: 0000:01:00.0 totalMemory: 8.00GiB freeMemory: 6.64GiB 2018-02-03 14:01:38.600763: I C:\tf_jenkins\workspace\rel-win\M\windows-gpu\PY\35\tensorflow\core\common_runtime\gpu\gpu_device.cc:1195] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 1070, pci bus id: 0000:01:00.0, compute capability: 6.1) I0203 14:01:40.828442 1860 tf_logging.py:110] Restoring parameters from ./snapshot/MoveToBeacon/fcn\model.pkl-100001 Traceback (most recent call last): File "C:\Users\Rick\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\client\session.py", line 1350, in _do_call return fn(*args) File "C:\Users\Rick\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\client\session.py", line 1329, in _run_fn status, run_metadata) File "C:\Users\Rick\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\framework\errors_impl.py", line 473, in exit c_api.TF_GetCode(self.status.status)) tensorflow.python.framework.errors_impl.InvalidArgumentError: Assign requires shapes of both tensors to match. lhs shape= [5,5,42,16] rhs shape= [5,5,23,16] [[Node: save/Assign_34 = Assign[T=DT_FLOAT, _class=["loc:@sconv1/weights"], use_locking=true, validate_shape=true, _device="/job:localhost/replica:0/task:0/device:CPU:0"](sconv1/weights/RMSProp, save/RestoreV2_34)]]

jkafrouni commented 6 years ago

@rwill128 did you figure out what was wrong ? running into the exact same problem now... (pysc2 version 1.2 from pip)