float32/float64 issue unresolved

[genekogan]

having same issue as described here even after setting floatX=float32. i have cuda 7.0, cudnn installed on ubuntu 14.04. skipthoughts is verified working fine. any idea what the issue could be?

$ THEANO_FLAGS='floatX=float32,device=gpu0,scan.allow_gc=True' python alignDraw.py models/coco-captions-32x32.json Using gpu device 0: GRID K520 (CNMeM is disabled, cuDNN 5005) alignDraw.py:342: FutureWarning: comparison to None will result in an elementwise object comparison in the future. if valData != None: Traceback (most recent call last): File "alignDraw.py", line 616, in rvae = ReccurentAttentionVAE(dimY, dimLangRNN, dimAlign, dimX, dimReadAttent, dimWriteAttent, dimRNNEnc, dimRNNDec, dimZ, runSteps, batch_size, reduceLRAfter, data, data_captions, valData=val_data, valDataCaptions=val_data_captions, pathToWeights=pathToWeights) File "alignDraw.py", line 354, in init self._kl_final, self._logpxz, self._log_likelihood, self._c_ts, self._c_ts_gener, self._x, self._y, self._run_steps, self._updates_train, self._updates_gener, self._read_attent_params, self._write_attent_params, self._write_attent_params_gener, self._alphas_gener, self._params, self._mu_prior_t_gener, self._log_sigma_prior_t_gener = build_lang_encoder_and_attention_vae_decoder(self.dimY, self.dimLangRNN, self.dimAlign, self.dimX, self.dimReadAttent, self.dimWriteAttent, self.dimRNNEnc, self.dimRNNDec, self.dimZ, self.runSteps, self.pathToWeights) File "alignDraw.py", line 293, in build_lang_encoder_and_attention_vae_decoder sequences=eps, outputs_info=[c0, h0_dec, cell0_dec, h0_enc, cell0_enc, kl_0, mu_prior_0, log_sigma_prior_0, None, None], non_sequences=all_params, n_steps=run_steps) File "/usr/local/lib/python2.7/dist-packages/theano/scan_module/scan.py", line 1041, in scan scan_outs = local_op(scan_inputs) File "/usr/local/lib/python2.7/dist-packages/theano/gof/op.py", line 611, in call node = self.make_node(_inputs, _kwargs) File "/usr/local/lib/python2.7/dist-packages/theano/scan_module/scan_op.py", line 538, in make_node inner_sitsot_out.type.dtype)) ValueError: When compiling the inner function of scan the following error has been encountered: The initial state (outputs_info in scan nomenclature) of variable IncSubtensor{Set;:int64:}.0 (argument number 1) has dtype float32, while the result of the inner function (fn) has dtype float64. This can happen if the inner function of scan results in an upcast or downcast.

[simplysimleprob]

I have exactly same issue. No luck so far :(

[strawberrysparkle]

I'm struggling with the same error as well, if anyone has been able to resolve this in the meantime since the previous folks posted. :)

Update: I was able to resolve this error by adding the Theano flag cast_policy=numpy+floatX. The command that worked for me was THEANO_FLAGS='scan.allow_gc=True,device=cuda,floatX=float32,cast_policy=numpy+floatX' python alignDraw.py. Hope this helps someone else too!

[YYlin]

when I python alignDraw.py THEANO_FLAGS='scan.allow_gc=True,device=cuda,floatX=float32,cast_policy=numpy+floatX', I find there is same problem could you tell me why? yulin@dlnlp01:~/gan_projects/text2image-master/coco$ python alignDraw.py models/coco-captions-32x32.json THEANO_FLAGS='scan.allow_gc=False,device=cuda,floatX=float32,cast_policy=numpy+floatX' /home/yulin/anaconda3/lib/python3.6/site-packages/h5py/init.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type. from ._conv import register_converters as _register_converters building gradient function 0:00:13.788787 building train function Traceback (most recent call last): File "alignDraw.py", line 628, in rvae.train(lr, epochs, save=save, validateAfter=validateAfter) File "alignDraw.py", line 495, in train self._build_train_function() File "alignDraw.py", line 445, in _build_train_function self._y: self.train_captions[self._index_cap, 0:self._cap_len] File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/function.py", line 317, in function output_keys=output_keys) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 449, in pfunc no_default_updates=no_default_updates) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 219, in rebuild_collect_shared cloned_v = clone_v_get_shared_updates(v, copy_inputs_over) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 93, in clone_v_get_shared_updates clone_v_get_shared_updates(i, copy_inputs_over) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 93, in clone_v_get_shared_updates clone_v_get_shared_updates(i, copy_inputs_over) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 93, in clone_v_get_shared_updates clone_v_get_shared_updates(i, copy_inputs_over) [Previous line repeated 1 more times] File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/compile/pfunc.py", line 96, in clone_v_get_shared_updates [clone_d[i] for i in owner.inputs], strict=rebuild_strict) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/gof/graph.py", line 242, in clone_with_new_inputs new_inputs[i] = curr.type.filter_variable(new) File "/home/yulin/anaconda3/lib/python3.6/site-packages/theano/tensor/type.py", line 234, in filter_variable self=self)) TypeError: Cannot convert Type TensorType(float32, matrix) (of Variable AdvancedSubtensor1.0) into Type TensorType(float64, matrix). You can try to manually convert AdvancedSubtensor1.0 into a TensorType(float64, matrix).