Re: r24323 - in /branches/disp_spin_speed: dep_check.py target_functions/relax_disp.py
Hi ed. I have already been looking at the systemtests. Our server has numpy 1.4.1 But some of the analytical models fails as well, since 1.4.1 does not recognise the 'out' keyword. Sigh... 2014-06-26 17:26 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi, I can look at it later. For requiring numpy 1.6, the following could be done: - If you could delete the printout in dep_check, that would be good. The printouts are only used for the essential packages before calling sys.exit(). If you are interested in a message to the user, the message needs to be modified to say something like numpy = 1.6 is required for..., rather than numpy einsum is not installed, and this needs to be used like the C_module_exp_fn_mesg variable in the dep_check and info modules rather than calling sys.stderr.write(). - Blacklist the system tests requiring this. This is absolutely essential for me to release the changes in a new relax version. Maybe you could tackle that one, as the problem is of your own doing ;) It's just a simple duplication of code in the setUp() method. Do you have a system with numpy 1.6 to test on? - A RelaxError in the relax_disp.select_model user function for just these models. Could you give a list of these models, as I don't know if all numeric models are affected, or if non-numeric models are affected. The best would be to create a new variable in specific_analyses.relax_disp.variables for this model list. - A GUI error blocking the execution of the analysis. This would require the model list in specific_analyses.relax_disp.variables to identify the problem. But I can look at this one. Cheers, Edward On 25 June 2014 20:14, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Edward. Do you fix this? Thanks. Best Troels 2014-06-25 20:06 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, For this change, it might be better to not output a message to sys.stderr. For relax users who are not interested in a dispersion analysis, there is no need to bother them. What I would suggest for safety is, in the backend for the relax_disp.select_model user function, that we raise a RelaxError if one of these models are selected but numpy.einsum does not exist. The RelaxError message could be that numpy version 1.6 or greater is required for the numeric dispersion models. A message could also appear in the GUI when the execute button is clicked on, if these models have been selected. That should be the best for the user experience when they have numpy 1.6. Regards, Edward On 25 June 2014 19:31, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Wed Jun 25 19:31:40 2014 New Revision: 24323 URL: http://svn.gna.org/viewcvs/relax?rev=24323view=rev Log: Implemented a dependency check for numpy einsum, so relax can still start. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/dep_check.py branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/dep_check.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/dep_check.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/dep_check.py (original) +++ branches/disp_spin_speed/dep_check.py Wed Jun 25 19:31:40 2014 @@ -57,6 +57,14 @@ # Optional packages. + +# numpy einsum import. +try: +from numpy import einsum +einsum_module = True +except ImportError: +sys.stderr.write(The dependency 'numpy einsum' has not been installed. Numerical models in dispersion analysis will not be possible.\n) +einsum_module = False # Bmrblib python package check. try: Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Wed Jun 25 19:31:40 2014 @@ -30,6 +30,7 @@ from numpy.ma import masked_equal # relax module imports. +import dep_check from lib.dispersion.b14 import r2eff_B14 from lib.dispersion.cr72 import r2eff_CR72 from lib.dispersion.dpl94 import r1rho_DPL94 @@ -40,13 +41,7 @@ from lib.dispersion.m61b import r1rho_M61b from lib.dispersion.mp05 import r1rho_MP05 from lib.dispersion.mmq_cr72 import r2eff_mmq_cr72 -from lib.dispersion.ns_cpmg_2site_3d import r2eff_ns_cpmg_2site_3D from lib.dispersion.ns_cpmg_2site_expanded import r2eff_ns_cpmg_2site_expanded -from lib.dispersion.ns_cpmg_2site_star import
Re: r24323 - in /branches/disp_spin_speed: dep_check.py target_functions/relax_disp.py
Hi Troels, This is ok. If you would like to have the relax numpy dependency increased to 1.6, then you'll just have to blacklist all system tests that do not work, so they obtain a printout at the end that X number of tests were skipped due to numpy 1.6. Then on the website and in the manual, we tell users to use numpy 1.6 at their own risk. We could even block the relaxation dispersion analysis in the pipe.create user function via a RelaxError if numpy 1.6, if all models are affected. But if this is the case, then maybe we make numpy = 1.6 an essential dependency which is checked in dep_check, calling sys.exit() if not. Regards, Edward On 27 June 2014 10:34, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi ed. I have already been looking at the systemtests. Our server has numpy 1.4.1 But some of the analytical models fails as well, since 1.4.1 does not recognise the 'out' keyword. Sigh... 2014-06-26 17:26 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi, I can look at it later. For requiring numpy 1.6, the following could be done: - If you could delete the printout in dep_check, that would be good. The printouts are only used for the essential packages before calling sys.exit(). If you are interested in a message to the user, the message needs to be modified to say something like numpy = 1.6 is required for..., rather than numpy einsum is not installed, and this needs to be used like the C_module_exp_fn_mesg variable in the dep_check and info modules rather than calling sys.stderr.write(). - Blacklist the system tests requiring this. This is absolutely essential for me to release the changes in a new relax version. Maybe you could tackle that one, as the problem is of your own doing ;) It's just a simple duplication of code in the setUp() method. Do you have a system with numpy 1.6 to test on? - A RelaxError in the relax_disp.select_model user function for just these models. Could you give a list of these models, as I don't know if all numeric models are affected, or if non-numeric models are affected. The best would be to create a new variable in specific_analyses.relax_disp.variables for this model list. - A GUI error blocking the execution of the analysis. This would require the model list in specific_analyses.relax_disp.variables to identify the problem. But I can look at this one. Cheers, Edward On 25 June 2014 20:14, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Edward. Do you fix this? Thanks. Best Troels 2014-06-25 20:06 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, For this change, it might be better to not output a message to sys.stderr. For relax users who are not interested in a dispersion analysis, there is no need to bother them. What I would suggest for safety is, in the backend for the relax_disp.select_model user function, that we raise a RelaxError if one of these models are selected but numpy.einsum does not exist. The RelaxError message could be that numpy version 1.6 or greater is required for the numeric dispersion models. A message could also appear in the GUI when the execute button is clicked on, if these models have been selected. That should be the best for the user experience when they have numpy 1.6. Regards, Edward On 25 June 2014 19:31, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Wed Jun 25 19:31:40 2014 New Revision: 24323 URL: http://svn.gna.org/viewcvs/relax?rev=24323view=rev Log: Implemented a dependency check for numpy einsum, so relax can still start. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/dep_check.py branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/dep_check.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/dep_check.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/dep_check.py (original) +++ branches/disp_spin_speed/dep_check.py Wed Jun 25 19:31:40 2014 @@ -57,6 +57,14 @@ # Optional packages. + +# numpy einsum import. +try: +from numpy import einsum +einsum_module = True +except ImportError: +sys.stderr.write(The dependency 'numpy einsum' has not been installed. Numerical models in dispersion analysis will not be possible.\n) +einsum_module = False # Bmrblib python package check. try: Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py
Skipping unit tests
Hej Edward. How do I skip unit tests? Best Troels ___ relax (http://www.nmr-relax.com) This is the relax-devel mailing list relax-devel@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel
Re: r24339 - /branches/disp_spin_speed/test_suite/system_tests/relax_disp.py
Wow, are their any dispersion models that can be used in this case? If not, we should probably add a specific_analyses.relax_disp.checks.check_numpy_out() function which raises a RelaxError when the out argument is unavailable. This function can then be called at the start of any relaxation dispersion specific user functions where it is important (minimise, calc, and grid_search for example). This will the prevent a user from encountering the numpy ValueError and clearly tell them that they need to upgrade their numpy version. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:29 2014 New Revision: 24339 URL: http://svn.gna.org/viewcvs/relax?rev=24339view=rev Log: Added a list of systemtests to skip, if numpy out argument is not available. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/test_suite/system_tests/relax_disp.py Modified: branches/disp_spin_speed/test_suite/system_tests/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/test_suite/system_tests/relax_disp.py?rev=24339r1=24338r2=24339view=diff == --- branches/disp_spin_speed/test_suite/system_tests/relax_disp.py (original) +++ branches/disp_spin_speed/test_suite/system_tests/relax_disp.py Fri Jun 27 12:58:29 2014 @@ -23,7 +23,7 @@ # Python module imports. from os import F_OK, access, getcwd, path, sep -from numpy import array, median +from numpy import array, median, version import re, math from tempfile import mkdtemp @@ -68,6 +68,73 @@ if methodName in to_skip: status.skipped_tests.append([methodName, 'Relax curve-fitting C module', self._skip_type]) +# Missing numpy out argument in numpy.multiply, numpy.add, numpy.subtract. +if not dep_check.numpy_out: +# The list of tests to skip. +to_skip = [ +test_baldwin_synthetic, +test_baldwin_synthetic_full, +test_bug_21460_disp_cluster_fail, +test_bug_22146_unpacking_r2a_r2b_cluster_B14, +test_bug_22146_unpacking_r2a_r2b_cluster_CR72, +test_bug_22146_unpacking_r2a_r2b_cluster_NS_3D, +test_bug_22146_unpacking_r2a_r2b_cluster_NS_STAR, +test_cpmg_synthetic_b14_to_ns3d_cluster, +test_cpmg_synthetic_b14_to_ns_star_cluster, +test_cpmg_synthetic_dx_map_points, +test_cpmg_synthetic_ns3d_to_b14, +test_cpmg_synthetic_ns3d_to_cr72, +test_cpmg_synthetic_ns3d_to_cr72_noise_cluster, +test_dpl94_data_to_dpl94, +test_hansen_cpmg_data_auto_analysis, +test_hansen_cpmg_data_auto_analysis_numeric, +test_hansen_cpmg_data_auto_analysis_r2eff, +test_hansen_cpmg_data_missing_auto_analysis, +test_hansen_cpmg_data_to_cr72, +test_hansen_cpmg_data_to_cr72_full, +test_hansen_cpmg_data_to_it99, +test_hansen_cpmg_data_to_lm63, +test_hansen_cpmg_data_to_ns_cpmg_2site_3D, +test_hansen_cpmg_data_to_ns_cpmg_2site_3D_full, +test_hansen_cpmg_data_to_ns_cpmg_2site_expanded, +test_hansen_cpmg_data_to_ns_cpmg_2site_star, +test_hansen_cpmg_data_to_ns_cpmg_2site_star_full, +test_korzhnev_2005_15n_dq_data, +test_korzhnev_2005_15n_mq_data, +test_korzhnev_2005_15n_sq_data, +test_korzhnev_2005_15n_zq_data, +test_korzhnev_2005_1h_mq_data, +test_korzhnev_2005_1h_sq_data, +test_korzhnev_2005_all_data, +test_korzhnev_2005_all_data_disp_speed_bug, +test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_cr72, + test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_cr72_full, + test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_tsmfk01, + test_kteilum_fmpoulsen_makke_cpmg_data_101m_guhcl_to_tsmfk01, +test_lm63_3site_synthetic, +test_m61_data_to_m61, +test_m61_exp_data_to_m61, +test_ns_mmq_3site, +test_ns_mmq_3site_linear, +test_ns_r1rho_3site, +test_ns_r1rho_3site_linear, +test_r1rho_kjaergaard_auto, +test_r1rho_kjaergaard_man, +test_r1rho_kjaergaard_missing_r1, +test_r20_from_min_r2eff_cpmg, +test_sod1wt_t25_to_cr72, +test_sprangers_data_to_mmq_cr72, +
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
Hi Ed. You can just grep for out or einsum in lib/dispersion. Best Troels 2014-06-27 14:15 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, I think if would be better if we caught this earlier, specifically the select_model() method of the specific_analyses.relax_disp.uf module if the user chooses a model which is not supported by their numpy version. Such a check with a RelaxError would only be two lines of code, plus a comment (which could include the current version in numpy.version.version). It would be very useful to add the models to a list variable in specific_analyses.relax_disp.variables so that I can replicate the checks in the GUI. Even better would be to add the 2 line check to a function in specific_analyses.relax_disp.checks to allow for consistent checking in all user interfaces. There could even be two separate tests, one for the numpy.einsum and one for the numpy out argument, each with their own model lists, if you wish. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:22 2014 New Revision: 24336 URL: http://svn.gna.org/viewcvs/relax?rev=24336view=rev Log: Added to target function that experiment_type_setup() should not be initiated, if numpy.einsum is missing. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24336r1=24335r2=24336view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Fri Jun 27 12:58:22 2014 @@ -365,8 +365,10 @@ # This is to make sure, that the chi2 values is not affected by missing values. self.mask_replace_blank = masked_equal(self.missing, 1.0) -# Check the experiment types, simplifying the data structures as needed. -self.experiment_type_setup() +# Check if eisum is available for numerical models. +if dep_check.einsum_module: +# Check the experiment types, simplifying the data structures as needed. +self.experiment_type_setup() # Scaling initialisation. self.scaling_flag = False ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list relax-comm...@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-commits ___ relax (http://www.nmr-relax.com) This is the relax-devel mailing list relax-devel@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel ___ relax (http://www.nmr-relax.com) This is the relax-devel mailing list relax-devel@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy = 1.6 is required.\n % numpy.version.version) +sys.exit() except ImportError: sys.stderr.write(The dependency 'numpy' has not been installed.\n) sys.exit() This might be the easiest way, just forcing users to upgrade. I've used the out argument a lot with the frame order analysis as well and plan to use it more often, so maybe we just have to put our foot down and make this a hard dependency. That will make everyone's life easier, and simplify what we as developers have to do to handle this. I can apply this to the trunk and release relax 3.2.3 with all the current fixes and this hard numpy version dependency, and then in a few weeks when you are ready, merge the disp_spin_speed branch and release relax 3.2.4. What do you think? Regards, Edward On 27 June 2014 14:40, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Ed. You can just grep for out or einsum in lib/dispersion. Best Troels 2014-06-27 14:15 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, I think if would be better if we caught this earlier, specifically the select_model() method of the specific_analyses.relax_disp.uf module if the user chooses a model which is not supported by their numpy version. Such a check with a RelaxError would only be two lines of code, plus a comment (which could include the current version in numpy.version.version). It would be very useful to add the models to a list variable in specific_analyses.relax_disp.variables so that I can replicate the checks in the GUI. Even better would be to add the 2 line check to a function in specific_analyses.relax_disp.checks to allow for consistent checking in all user interfaces. There could even be two separate tests, one for the numpy.einsum and one for the numpy out argument, each with their own model lists, if you wish. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:22 2014 New Revision: 24336 URL: http://svn.gna.org/viewcvs/relax?rev=24336view=rev Log: Added to target function that experiment_type_setup() should not be initiated, if numpy.einsum is missing. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24336r1=24335r2=24336view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Fri Jun 27 12:58:22 2014 @@ -365,8 +365,10 @@ # This is to make sure, that the chi2 values is not affected by missing values. self.mask_replace_blank = masked_equal(self.missing, 1.0) -# Check the experiment types, simplifying the data structures as needed. -self.experiment_type_setup() +# Check if eisum is available for numerical models. +if dep_check.einsum_module: +# Check the experiment types, simplifying the data structures as needed. +self.experiment_type_setup() # Scaling initialisation. self.scaling_flag = False ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list relax-comm...@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-commits
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy = 1.6 is required.\n % numpy.version.version) +sys.exit() except ImportError: sys.stderr.write(The dependency 'numpy' has not been installed.\n) sys.exit() This might be the easiest way, just forcing users to upgrade. I've used the out argument a lot with the frame order analysis as well and plan to use it more often, so maybe we just have to put our foot down and make this a hard dependency. That will make everyone's life easier, and simplify what we as developers have to do to handle this. I can apply this to the trunk and release relax 3.2.3 with all the current fixes and this hard numpy version dependency, and then in a few weeks when you are ready, merge the disp_spin_speed branch and release relax 3.2.4. What do you think? Regards, Edward On 27 June 2014 14:40, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Ed. You can just grep for out or einsum in lib/dispersion. Best Troels 2014-06-27 14:15 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, I think if would be better if we caught this earlier, specifically the select_model() method of the specific_analyses.relax_disp.uf module if the user chooses a model which is not supported by their numpy version. Such a check with a RelaxError would only be two lines of code, plus a comment (which could include the current version in numpy.version.version). It would be very useful to add the models to a list variable in specific_analyses.relax_disp.variables so that I can replicate the checks in the GUI. Even better would be to add the 2 line check to a function in specific_analyses.relax_disp.checks to allow for consistent checking in all user interfaces. There could even be two separate tests, one for the numpy.einsum and one for the numpy out argument, each with their own model lists, if you wish. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:22 2014 New Revision: 24336 URL: http://svn.gna.org/viewcvs/relax?rev=24336view=rev Log: Added to target function that experiment_type_setup() should not be initiated, if numpy.einsum is missing. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24336r1=24335r2=24336view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Fri Jun 27 12:58:22 2014 @@ -365,8 +365,10 @@ # This is to make sure, that the chi2 values is not affected by missing values. self.mask_replace_blank = masked_equal(self.missing, 1.0) -# Check the experiment types, simplifying the data structures as needed. -self.experiment_type_setup() +# Check if eisum is available for numerical models. +if dep_check.einsum_module: +# Check the experiment types, simplifying the data structures as needed. +self.experiment_type_setup() # Scaling initialisation. self.scaling_flag = False ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
Let me see. You've covered most of the running list, there's just one or two things left. I'll add a few more trivial things as well. - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. I'll keep looking. I'm guessing you will not be tackling the numeric models to speed them up to match 'NS CPMG 2-site expanded' model via the ultimate solution of brute-force expansion (http://www.mail-archive.com/relax-users@gna.org/msg01641.html) any time soon. This will have a similar effect as the ultimate speed up at http://www.mail-archive.com/relax-devel%40gna.org/msg05691.html which you have so successfully achieved in this branch. Can you remember if there were any other changes required? I will create a new timings file just before tagging the relax release. Cheers, Edward On 27 June 2014 16:53, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy = 1.6 is required.\n % numpy.version.version) +sys.exit() except ImportError: sys.stderr.write(The dependency 'numpy' has not been installed.\n) sys.exit() This might be the easiest way, just forcing users to upgrade. I've used the out argument a lot with the frame order analysis as well and plan to use it more often, so maybe we just have to put our foot down and make this a hard dependency. That will make everyone's life easier, and simplify what we as developers have to do to handle this. I can apply this to the trunk and release relax 3.2.3 with all the current fixes and this hard numpy version dependency, and then in a few weeks when you are ready, merge the disp_spin_speed branch and release relax 3.2.4. What do you think? Regards, Edward On 27 June 2014 14:40, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Ed. You can just grep for out or einsum in lib/dispersion. Best Troels 2014-06-27 14:15 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, I think if would be better if we caught this earlier, specifically the select_model() method of the specific_analyses.relax_disp.uf module if the user chooses a model which is not supported by their numpy version. Such a check with a RelaxError would only be two lines of code, plus a comment (which could include the current version in numpy.version.version). It would be very useful to add the models to a list variable in specific_analyses.relax_disp.variables so that I can replicate the checks in the GUI. Even better would be to add the 2 line check to a function in specific_analyses.relax_disp.checks to allow for consistent checking in all user interfaces. There could even be two separate tests, one for the numpy.einsum and one for the numpy out argument, each with their own model lists, if you wish. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author:
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
To add to the list of trivialities to fix: - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. Cheers, Edward On 27 June 2014 17:23, Edward d'Auvergne edw...@nmr-relax.com wrote: Let me see. You've covered most of the running list, there's just one or two things left. I'll add a few more trivial things as well. - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. I'll keep looking. I'm guessing you will not be tackling the numeric models to speed them up to match 'NS CPMG 2-site expanded' model via the ultimate solution of brute-force expansion (http://www.mail-archive.com/relax-users@gna.org/msg01641.html) any time soon. This will have a similar effect as the ultimate speed up at http://www.mail-archive.com/relax-devel%40gna.org/msg05691.html which you have so successfully achieved in this branch. Can you remember if there were any other changes required? I will create a new timings file just before tagging the relax release. Cheers, Edward On 27 June 2014 16:53, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy = 1.6 is required.\n % numpy.version.version) +sys.exit() except ImportError: sys.stderr.write(The dependency 'numpy' has not been installed.\n) sys.exit() This might be the easiest way, just forcing users to upgrade. I've used the out argument a lot with the frame order analysis as well and plan to use it more often, so maybe we just have to put our foot down and make this a hard dependency. That will make everyone's life easier, and simplify what we as developers have to do to handle this. I can apply this to the trunk and release relax 3.2.3 with all the current fixes and this hard numpy version dependency, and then in a few weeks when you are ready, merge the disp_spin_speed branch and release relax 3.2.4. What do you think? Regards, Edward On 27 June 2014 14:40, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: Hi Ed. You can just grep for out or einsum in lib/dispersion. Best Troels 2014-06-27 14:15 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hi Troels, I think if would be better if we caught this earlier, specifically the select_model() method of the specific_analyses.relax_disp.uf module if the user chooses a model which is not supported by their numpy version. Such a check with a RelaxError would only be two lines of code, plus a comment (which could include the current version in numpy.version.version). It would be very useful to add the models to a list variable in specific_analyses.relax_disp.variables so that I can replicate the checks in the GUI. Even better would be to add the 2 line check to a function in specific_analyses.relax_disp.checks to allow for consistent checking in all user interfaces. There could
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
Also the documentation in the get_back_calc() method of the target function needs a little polish - it is not a float that is returned. So, the running list: - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. - The get_back_calc() method of the target function needs a little polish - it is not a float that is returned. Cheers, Edward On 27 June 2014 17:26, Edward d'Auvergne edw...@nmr-relax.com wrote: To add to the list of trivialities to fix: - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. Cheers, Edward On 27 June 2014 17:23, Edward d'Auvergne edw...@nmr-relax.com wrote: Let me see. You've covered most of the running list, there's just one or two things left. I'll add a few more trivial things as well. - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. I'll keep looking. I'm guessing you will not be tackling the numeric models to speed them up to match 'NS CPMG 2-site expanded' model via the ultimate solution of brute-force expansion (http://www.mail-archive.com/relax-users@gna.org/msg01641.html) any time soon. This will have a similar effect as the ultimate speed up at http://www.mail-archive.com/relax-devel%40gna.org/msg05691.html which you have so successfully achieved in this branch. Can you remember if there were any other changes required? I will create a new timings file just before tagging the relax release. Cheers, Edward On 27 June 2014 16:53, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy = 1.6 is required.\n % numpy.version.version) +sys.exit() except ImportError: sys.stderr.write(The dependency 'numpy' has not been installed.\n) sys.exit() This might be the easiest way, just forcing users to upgrade. I've used the out argument a lot with the frame order analysis as well and plan to use it more often, so maybe we just have to put our foot down and make this a hard dependency. That will make everyone's life easier, and simplify what we as developers have to
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
There are also a number of double spacing issues. You can find these by running: $ ./devel_scripts/code_validator lib/dispersion/b14.py $ ./devel_scripts/code_validator lib/dispersion/cr72.py Etc. Some things can be ignored, like the capitals in the function name and double spacing or trailing spacing for equations in docstrings. This script will identify most formatting problems remaining. Cheers, Edward On 27 June 2014 17:28, Edward d'Auvergne edw...@nmr-relax.com wrote: Also the documentation in the get_back_calc() method of the target function needs a little polish - it is not a float that is returned. So, the running list: - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. - The get_back_calc() method of the target function needs a little polish - it is not a float that is returned. Cheers, Edward On 27 June 2014 17:26, Edward d'Auvergne edw...@nmr-relax.com wrote: To add to the list of trivialities to fix: - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. Cheers, Edward On 27 June 2014 17:23, Edward d'Auvergne edw...@nmr-relax.com wrote: Let me see. You've covered most of the running list, there's just one or two things left. I'll add a few more trivial things as well. - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. I'll keep looking. I'm guessing you will not be tackling the numeric models to speed them up to match 'NS CPMG 2-site expanded' model via the ultimate solution of brute-force expansion (http://www.mail-archive.com/relax-users@gna.org/msg01641.html) any time soon. This will have a similar effect as the ultimate speed up at http://www.mail-archive.com/relax-devel%40gna.org/msg05691.html which you have so successfully achieved in this branch. Can you remember if there were any other changes required? I will create a new timings file just before tagging the relax release. Cheers, Edward On 27 June 2014 16:53, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0 lib/dispersion/tp02.py:0 lib/dispersion/tsmfk01.py:0 lib/dispersion/two_point.py:0 What do you think of the idea of making this a hard dependency: Index: dep_check.py === --- dep_check.py(revision 24343) +++ dep_check.py(working copy) @@ -37,6 +37,9 @@ # numpy. try: import numpy +if float(numpy.version.version[:3]) 1.6: +sys.stderr.write(Version %s of the 'numpy' dependency is not supported, numpy =
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
Here is one more point. For the DPL94 model, the denominator is kex**2 + spin_lock_fields**2. This can only be zero when kex is zero, and then there should be no exchange. Therefore we can add a kex == 0 check at the start to return R1_R2 and then delete the mask_denom_zero logic. This should slightly speed things up by removing one 'if' statement. Also, kex**2 is not a repetitive operation. Cheers, Edward On 27 June 2014 17:33, Edward d'Auvergne edw...@nmr-relax.com wrote: There are also a number of double spacing issues. You can find these by running: $ ./devel_scripts/code_validator lib/dispersion/b14.py $ ./devel_scripts/code_validator lib/dispersion/cr72.py Etc. Some things can be ignored, like the capitals in the function name and double spacing or trailing spacing for equations in docstrings. This script will identify most formatting problems remaining. Cheers, Edward On 27 June 2014 17:28, Edward d'Auvergne edw...@nmr-relax.com wrote: Also the documentation in the get_back_calc() method of the target function needs a little polish - it is not a float that is returned. So, the running list: - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. - The get_back_calc() method of the target function needs a little polish - it is not a float that is returned. Cheers, Edward On 27 June 2014 17:26, Edward d'Auvergne edw...@nmr-relax.com wrote: To add to the list of trivialities to fix: - The 'chi2_sum' initialisation in func_ns_mmq_2site() should go. Cheers, Edward On 27 June 2014 17:23, Edward d'Auvergne edw...@nmr-relax.com wrote: Let me see. You've covered most of the running list, there's just one or two things left. I'll add a few more trivial things as well. - Some modules have = operators outside of functions that need spaces around them. - Trailing whitespace, most of it is in the trunk, but there are some new ones. Just run: grep $ lib/dispersion/* | grep -v - There are some empty lines straight after a 'for' loop in many places. - Your copyright in the tsmfk01.py file needs to be extended to 2014. - There are some FIXME comments in the lib.dispersion package. - In some places, comment lines require empty lines before them (in target_functions.relax_disp and lib.dispersion). - Not very important, but the target function calc_ns_mmq_3site_chi2() can have the Once off parameter conversions shifted into the lib.dispersion module to simplify this. - The newline spacing between target function methods needs some fixes. I'll keep looking. I'm guessing you will not be tackling the numeric models to speed them up to match 'NS CPMG 2-site expanded' model via the ultimate solution of brute-force expansion (http://www.mail-archive.com/relax-users@gna.org/msg01641.html) any time soon. This will have a similar effect as the ultimate speed up at http://www.mail-archive.com/relax-devel%40gna.org/msg05691.html which you have so successfully achieved in this branch. Can you remember if there were any other changes required? I will create a new timings file just before tagging the relax release. Cheers, Edward On 27 June 2014 16:53, Troels Emtekær Linnet tlin...@nmr-relax.com wrote: This sounds good. What is needed to be done, to merge disp_spin_speed in trunk now? Best Troels 2014-06-27 15:53 GMT+02:00 Edward d'Auvergne edw...@nmr-relax.com: Hmmm: $ grep -c out=\|einsum lib/dispersion/* lib/dispersion/b14.py:0 lib/dispersion/cr72.py:2 lib/dispersion/dpl94.py:0 lib/dispersion/__init__.py:0 lib/dispersion/it99.py:0 lib/dispersion/lm63_3site.py:0 lib/dispersion/lm63.py:0 lib/dispersion/m61b.py:0 lib/dispersion/m61.py:0 lib/dispersion/matrix_exponential.py:5 lib/dispersion/matrix_power.py:0 lib/dispersion/mmq_cr72.py:0 lib/dispersion/mp05.py:0 lib/dispersion/ns_cpmg_2site_3d.py:6 lib/dispersion/ns_cpmg_2site_expanded.py:0 lib/dispersion/ns_cpmg_2site_star.py:3 lib/dispersion/ns_matrices.py:0 lib/dispersion/ns_mmq_2site.py:12 lib/dispersion/ns_mmq_3site.py:12 lib/dispersion/ns_r1rho_2site.py:3 lib/dispersion/ns_r1rho_3site.py:3 lib/dispersion/tap03.py:0
Re: r24323 - in /branches/disp_spin_speed: dep_check.py target_functions/relax_disp.py
Hi Troels, Because of the change http://www.mail-archive.com/relax-commits@gna.org/msg21976.html, the changes of r24323 are now redundant so you can revert them. Cheers, Edward On 25 June 2014 19:31, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Wed Jun 25 19:31:40 2014 New Revision: 24323 URL: http://svn.gna.org/viewcvs/relax?rev=24323view=rev Log: Implemented a dependency check for numpy einsum, so relax can still start. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/dep_check.py branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/dep_check.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/dep_check.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/dep_check.py (original) +++ branches/disp_spin_speed/dep_check.py Wed Jun 25 19:31:40 2014 @@ -57,6 +57,14 @@ # Optional packages. + +# numpy einsum import. +try: +from numpy import einsum +einsum_module = True +except ImportError: +sys.stderr.write(The dependency 'numpy einsum' has not been installed. Numerical models in dispersion analysis will not be possible.\n) +einsum_module = False # Bmrblib python package check. try: Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24323r1=24322r2=24323view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Wed Jun 25 19:31:40 2014 @@ -30,6 +30,7 @@ from numpy.ma import masked_equal # relax module imports. +import dep_check from lib.dispersion.b14 import r2eff_B14 from lib.dispersion.cr72 import r2eff_CR72 from lib.dispersion.dpl94 import r1rho_DPL94 @@ -40,13 +41,7 @@ from lib.dispersion.m61b import r1rho_M61b from lib.dispersion.mp05 import r1rho_MP05 from lib.dispersion.mmq_cr72 import r2eff_mmq_cr72 -from lib.dispersion.ns_cpmg_2site_3d import r2eff_ns_cpmg_2site_3D from lib.dispersion.ns_cpmg_2site_expanded import r2eff_ns_cpmg_2site_expanded -from lib.dispersion.ns_cpmg_2site_star import r2eff_ns_cpmg_2site_star -from lib.dispersion.ns_mmq_3site import r2eff_ns_mmq_3site_mq, r2eff_ns_mmq_3site_sq_dq_zq -from lib.dispersion.ns_mmq_2site import r2eff_ns_mmq_2site_mq, r2eff_ns_mmq_2site_sq_dq_zq -from lib.dispersion.ns_r1rho_2site import ns_r1rho_2site -from lib.dispersion.ns_r1rho_3site import ns_r1rho_3site from lib.dispersion.ns_matrices import r180x_3d from lib.dispersion.tp02 import r1rho_TP02 from lib.dispersion.tap03 import r1rho_TAP03 @@ -56,6 +51,14 @@ from target_functions.chi2 import chi2, chi2_rankN from specific_analyses.relax_disp.variables import EXP_TYPE_CPMG_DQ, EXP_TYPE_CPMG_MQ, EXP_TYPE_CPMG_PROTON_MQ, EXP_TYPE_CPMG_PROTON_SQ, EXP_TYPE_CPMG_SQ, EXP_TYPE_CPMG_ZQ, EXP_TYPE_LIST_CPMG, EXP_TYPE_R1RHO, MODEL_B14, MODEL_B14_FULL, MODEL_CR72, MODEL_CR72_FULL, MODEL_DPL94, MODEL_IT99, MODEL_LIST_CPMG, MODEL_LIST_CPMG_FULL, MODEL_LIST_FULL, MODEL_LIST_MMQ, MODEL_LIST_MQ_CPMG, MODEL_LIST_R1RHO, MODEL_LIST_R1RHO_FULL, MODEL_LM63, MODEL_LM63_3SITE, MODEL_M61, MODEL_M61B, MODEL_MP05, MODEL_MMQ_CR72, MODEL_NOREX, MODEL_NS_CPMG_2SITE_3D, MODEL_NS_CPMG_2SITE_3D_FULL, MODEL_NS_CPMG_2SITE_EXPANDED, MODEL_NS_CPMG_2SITE_STAR, MODEL_NS_CPMG_2SITE_STAR_FULL, MODEL_NS_MMQ_2SITE, MODEL_NS_MMQ_3SITE, MODEL_NS_MMQ_3SITE_LINEAR, MODEL_NS_R1RHO_2SITE, MODEL_NS_R1RHO_3SITE, MODEL_NS_R1RHO_3SITE_LINEAR, MODEL_TAP03, MODEL_TP02, MODEL_TSMFK01 +# Check if eisum is available for numerical models. +if dep_check.einsum_module: +from lib.dispersion.ns_cpmg_2site_3d import r2eff_ns_cpmg_2site_3D +from lib.dispersion.ns_cpmg_2site_star import r2eff_ns_cpmg_2site_star +from lib.dispersion.ns_mmq_3site import r2eff_ns_mmq_3site_mq, r2eff_ns_mmq_3site_sq_dq_zq +from lib.dispersion.ns_mmq_2site import r2eff_ns_mmq_2site_mq, r2eff_ns_mmq_2site_sq_dq_zq +from lib.dispersion.ns_r1rho_2site import ns_r1rho_2site +from lib.dispersion.ns_r1rho_3site import ns_r1rho_3site class Dispersion: def __init__(self, model=None, num_params=None, num_spins=None, num_frq=None, exp_types=None, values=None, errors=None, missing=None, frqs=None, frqs_H=None, cpmg_frqs=None, spin_lock_nu1=None, chemical_shifts=None, offset=None, tilt_angles=None, r1=None, relax_times=None, scaling_matrix=None, recalc_tau=True): ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list relax-comm...@gna.org To unsubscribe from this list, get a
Re: r24339 - /branches/disp_spin_speed/test_suite/system_tests/relax_disp.py
Hi Troels, Again because of the strict numpy = 1.6 requirement now found in relax, this change is also not required and can be reverted. Cheers, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:29 2014 New Revision: 24339 URL: http://svn.gna.org/viewcvs/relax?rev=24339view=rev Log: Added a list of systemtests to skip, if numpy out argument is not available. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/test_suite/system_tests/relax_disp.py Modified: branches/disp_spin_speed/test_suite/system_tests/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/test_suite/system_tests/relax_disp.py?rev=24339r1=24338r2=24339view=diff == --- branches/disp_spin_speed/test_suite/system_tests/relax_disp.py (original) +++ branches/disp_spin_speed/test_suite/system_tests/relax_disp.py Fri Jun 27 12:58:29 2014 @@ -23,7 +23,7 @@ # Python module imports. from os import F_OK, access, getcwd, path, sep -from numpy import array, median +from numpy import array, median, version import re, math from tempfile import mkdtemp @@ -68,6 +68,73 @@ if methodName in to_skip: status.skipped_tests.append([methodName, 'Relax curve-fitting C module', self._skip_type]) +# Missing numpy out argument in numpy.multiply, numpy.add, numpy.subtract. +if not dep_check.numpy_out: +# The list of tests to skip. +to_skip = [ +test_baldwin_synthetic, +test_baldwin_synthetic_full, +test_bug_21460_disp_cluster_fail, +test_bug_22146_unpacking_r2a_r2b_cluster_B14, +test_bug_22146_unpacking_r2a_r2b_cluster_CR72, +test_bug_22146_unpacking_r2a_r2b_cluster_NS_3D, +test_bug_22146_unpacking_r2a_r2b_cluster_NS_STAR, +test_cpmg_synthetic_b14_to_ns3d_cluster, +test_cpmg_synthetic_b14_to_ns_star_cluster, +test_cpmg_synthetic_dx_map_points, +test_cpmg_synthetic_ns3d_to_b14, +test_cpmg_synthetic_ns3d_to_cr72, +test_cpmg_synthetic_ns3d_to_cr72_noise_cluster, +test_dpl94_data_to_dpl94, +test_hansen_cpmg_data_auto_analysis, +test_hansen_cpmg_data_auto_analysis_numeric, +test_hansen_cpmg_data_auto_analysis_r2eff, +test_hansen_cpmg_data_missing_auto_analysis, +test_hansen_cpmg_data_to_cr72, +test_hansen_cpmg_data_to_cr72_full, +test_hansen_cpmg_data_to_it99, +test_hansen_cpmg_data_to_lm63, +test_hansen_cpmg_data_to_ns_cpmg_2site_3D, +test_hansen_cpmg_data_to_ns_cpmg_2site_3D_full, +test_hansen_cpmg_data_to_ns_cpmg_2site_expanded, +test_hansen_cpmg_data_to_ns_cpmg_2site_star, +test_hansen_cpmg_data_to_ns_cpmg_2site_star_full, +test_korzhnev_2005_15n_dq_data, +test_korzhnev_2005_15n_mq_data, +test_korzhnev_2005_15n_sq_data, +test_korzhnev_2005_15n_zq_data, +test_korzhnev_2005_1h_mq_data, +test_korzhnev_2005_1h_sq_data, +test_korzhnev_2005_all_data, +test_korzhnev_2005_all_data_disp_speed_bug, +test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_cr72, + test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_cr72_full, + test_kteilum_fmpoulsen_makke_cpmg_data_048m_guhcl_to_tsmfk01, + test_kteilum_fmpoulsen_makke_cpmg_data_101m_guhcl_to_tsmfk01, +test_lm63_3site_synthetic, +test_m61_data_to_m61, +test_m61_exp_data_to_m61, +test_ns_mmq_3site, +test_ns_mmq_3site_linear, +test_ns_r1rho_3site, +test_ns_r1rho_3site_linear, +test_r1rho_kjaergaard_auto, +test_r1rho_kjaergaard_man, +test_r1rho_kjaergaard_missing_r1, +test_r20_from_min_r2eff_cpmg, +test_sod1wt_t25_to_cr72, +test_sprangers_data_to_mmq_cr72, +test_sprangers_data_to_ns_mmq_2site, +test_tp02_data_to_mp05, +test_tp02_data_to_ns_r1rho_2site, +test_tp02_data_to_ns_r1rho_2site_cluster, +test_tp02_data_to_tap03, +test_tp02_data_to_tp02 +] + +# Store in the status object. +if methodName in to_skip: +
Re: r24336 - /branches/disp_spin_speed/target_functions/relax_disp.py
This change is also redundant thanks to http://www.mail-archive.com/relax-commits@gna.org/msg21976.html. Regards, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:22 2014 New Revision: 24336 URL: http://svn.gna.org/viewcvs/relax?rev=24336view=rev Log: Added to target function that experiment_type_setup() should not be initiated, if numpy.einsum is missing. Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/target_functions/relax_disp.py Modified: branches/disp_spin_speed/target_functions/relax_disp.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/target_functions/relax_disp.py?rev=24336r1=24335r2=24336view=diff == --- branches/disp_spin_speed/target_functions/relax_disp.py (original) +++ branches/disp_spin_speed/target_functions/relax_disp.py Fri Jun 27 12:58:22 2014 @@ -365,8 +365,10 @@ # This is to make sure, that the chi2 values is not affected by missing values. self.mask_replace_blank = masked_equal(self.missing, 1.0) -# Check the experiment types, simplifying the data structures as needed. -self.experiment_type_setup() +# Check if eisum is available for numerical models. +if dep_check.einsum_module: +# Check the experiment types, simplifying the data structures as needed. +self.experiment_type_setup() # Scaling initialisation. self.scaling_flag = False ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list relax-comm...@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-commits ___ relax (http://www.nmr-relax.com) This is the relax-devel mailing list relax-devel@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel
Re: r24337 - /branches/disp_spin_speed/dep_check.py
Hi Troels, This commit is also no longer required because of http://www.mail-archive.com/relax-commits@gna.org/msg21976.html. Feel free to revert. Regards, Edward On 27 June 2014 12:58, tlin...@nmr-relax.com wrote: Author: tlinnet Date: Fri Jun 27 12:58:24 2014 New Revision: 24337 URL: http://svn.gna.org/viewcvs/relax?rev=24337view=rev Log: Added a numpy out argument check to dep_check.py. This is for checking the out argument of: numpy.multiply numpy.add numpy.subtract Task #7807 (https://gna.org/task/index.php?7807): Speed-up of dispersion models for Clustered analysis. Modified: branches/disp_spin_speed/dep_check.py Modified: branches/disp_spin_speed/dep_check.py URL: http://svn.gna.org/viewcvs/relax/branches/disp_spin_speed/dep_check.py?rev=24337r1=24336r2=24337view=diff == --- branches/disp_spin_speed/dep_check.py (original) +++ branches/disp_spin_speed/dep_check.py Fri Jun 27 12:58:24 2014 @@ -64,6 +64,16 @@ einsum_module = True except ImportError: einsum_module = False + +# numpy out argument to numpy.multiply or numpy.add. +try: +a = numpy.array([1]) +numpy.multiply(1, 1, a) +numpy.add(1, 1, a) +numpy.subtract(1, 1, a) +numpy_out = True +except ValueError: +numpy_out = False # Bmrblib python package check. try: ___ relax (http://www.nmr-relax.com) This is the relax-commits mailing list relax-comm...@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-commits ___ relax (http://www.nmr-relax.com) This is the relax-devel mailing list relax-devel@gna.org To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-devel
