Error in Running Example

[yecyn]

Hello all,

I was trying to run the example 1 on Google Collab: https://colab.research.google.com/drive/1Dd289fKUinhtnG4OU8ghUwjidYWtc5PY?usp=sharing

But I run into an error, and I was wondering if anyone has run into the same error or if anyone knows how to fix it. Thank you so much! I really appreciate it!

---

TypeError Traceback (most recent call last) in 1 from aifeynman.S_run_aifeynman import run_aifeynman 2 # Run example 1 as the regression dataset ----> 3 run_aifeynman("/content/AI-Feynman/example_data/","example1.txt",30,"14ops.txt", polyfit_deg=3, NN_epochs=400)

2 frames /usr/local/lib/python3.7/dist-packages/torch/_tensor.py in array(self, dtype) 755 return handle_torch_function(Tensor.array, (self,), self, dtype=dtype) 756 if dtype is None: --> 757 return self.numpy() 758 else: 759 return self.numpy().astype(dtype, copy=False)

TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.

[kvyaswanth]

It got this error here. I just used .cpu() for every element and it was resolved.

[yecyn]

I changed it as follows:

else: idx_min = np.argmin(np.array([symmetry_plus_result[0].cpu(), symmetry_minus_result[0].cpu(), symmetry_multiply_result[0].cpu(), symmetry_divide_result[0].cpu(), separability_plus_result[0].cpu(), separability_multiply_result[0].cpu()]))

But I'm still getting the same error.

[kvyaswanth]

can you paste the error output here

[yecyn]

I tried it again this morning, and it worked! Thank you so much!!

[yecyn]

Hi, I was using the same code on a different dataset, but i ended up getting another error on the same line as before, I was wondering if you know how to fix it? Thank you so much!

---

AttributeError Traceback (most recent call last) in 1 from aifeynman.S_run_aifeynman import run_aifeynman 2 # Run Bass Example as the regression dataset ----> 3 run_aifeynman("/content/","Diamonds_Clean.txt",30,"14ops.txt", polyfit_deg=3, NN_epochs=400)

1 frames /content/AI-Feynman/aifeynman/S_run_aifeynman.py in run_AI_all(pathdir, filename, BF_try_time, BF_ops_file_type, polyfit_deg, NN_epochs, PA) 94 idx_min = -1 95 else: ---> 96 idx_min = np.argmin(np.array([symmetry_plus_result[0].cpu(), symmetry_minus_result[0].cpu(), symmetry_multiply_result[0].cpu(), symmetry_divide_result[0].cpu(), separability_plus_result[0].cpu(), separability_multiply_result[0].cpu()])) 97 print("") 98 # Check if compositionality is better than the best so far

AttributeError: 'int' object has no attribute 'cpu'

I have also tried the code without adding .cpu(), and it still has the same errors:

---

AttributeError Traceback (most recent call last) in 1 from aifeynman.S_run_aifeynman import run_aifeynman 2 # Run Bass Example as the regression dataset ----> 3 run_aifeynman("/content/","Diamonds_Clean.txt",30,"14ops.txt", polyfit_deg=3, NN_epochs=400)

1 frames /content/AI-Feynman/aifeynman/S_run_aifeynman.py in run_AI_all(pathdir, filename, BF_try_time, BF_ops_file_type, polyfit_deg, NN_epochs, PA) 94 idx_min = -1 95 else: ---> 96 idx_min = np.argmin(np.array([symmetry_plus_result[0], symmetry_minus_result[0], symmetry_multiply_result[0], symmetry_divide_result[0], separability_plus_result[0], separability_multiply_result[0]])) 97 print("") 98 # Check if compositionality is better than the best so far

AttributeError: 'int' object has no attribute 'cpu'

[kvyaswanth]

the check symmetry and separability functions either return a tensor or -1

in this case you can remove .cpu(), since its not returning a tensor

[yecyn]

I tried it again without .cpu(), and it worked. But I'm getting another error:

I double checked my data, and it does not have any NaN or inf values.

---

LinAlgError Traceback (most recent call last) in 1 from aifeynman.S_run_aifeynman import run_aifeynman 2 # Run Bass Example as the regression dataset ----> 3 run_aifeynman("/content/","Diamonds_Clean.txt",30,"14ops.txt", polyfit_deg=3, NN_epochs=400)

6 frames /content/AI-Feynman/aifeynman/S_run_aifeynman.py in run_aifeynman(pathdir, filename, BF_try_time, BF_ops_file_type, polyfit_deg, NN_epochs, vars_name, test_percentage) 272 PA = ParetoSet() 273 # Run the code on the train data --> 274 PA = run_AI_all(pathdir,filename+"_train",BF_try_time,BF_ops_file_type, polyfit_deg, NN_epochs, PA=PA) 275 PA_list = PA.get_pareto_points() 276

/content/AI-Feynman/aifeynman/S_run_aifeynman.py in run_AI_all(pathdir, filename, BF_try_time, BF_ops_file_type, polyfit_deg, NN_epochs, PA) 152 if len(data[0])>3: 153 # find the best separability indices --> 154 decomp_idx = identify_decompositions(pathdir,filename, model_feynman) 155 brute_force_gen_sym("results/","gradients_gensym%s" %filename,600,"14ops.txt") 156 bf_all_output = np.loadtxt("results_gen_sym.dat", dtype="str")

/content/AI-Feynman/aifeynman/S_gradient_decomposition.py in identify_decompositions(pathdir, filename, model, max_subset_size, visualize) 256 y = torch.Tensor(data[:, [-1]]) 257 # Return best decomposition --> 258 all_scores = filter_decompositions_relative_scoring(X, y, model, visualize=visualize) 259 assert(all_scores) 260 best_decomposition = all_scores[0][1]

/content/AI-Feynman/aifeynman/S_gradient_decomposition.py in filter_decompositions_relative_scoring(X, y, model, max_subset_size, visualize) 198 for i in range(random_indices.shape[0]): 199 samples = draw_samples(X, y, model, s, NUM_SAMPLES, point=randomindices[i]) --> 200 score, = score_consistency(evaluate_derivatives_andrew(model, s, samples)) 201 hypot_scores.append(score) 202 for i in range(random_indices.shape[0]):

/content/AI-Feynman/aifeynman/S_gradient_decomposition.py in score_consistency(grads_tensor) 109 A = np.array(normalized_grads) 110 D = np.einsum('ij,ik', A, A) --> 111 evals, evecs = np.linalg.eig(D) 112 nv = evals.shape[0] 113 assert(nv == A.shape[1])

<__array_function__ internals> in eig(*args, **kwargs) [/usr/local/lib/python3.7/dist-packages/numpy/linalg/linalg.py](https://localhost:8080/#) in eig(a) 1315 _assert_stacked_2d(a) 1316 _assert_stacked_square(a) -> 1317 _assert_finite(a) 1318 t, result_t = _commonType(a) 1319 [/usr/local/lib/python3.7/dist-packages/numpy/linalg/linalg.py](https://localhost:8080/#) in _assert_finite(*arrays) 206 for a in arrays: 207 if not isfinite(a).all(): --> 208 raise LinAlgError("Array must not contain infs or NaNs") 209 210 def _is_empty_2d(arr): LinAlgError: Array must not contain infs or NaNs