diff --git a/.ipynb_checkpoints/raremodel-nb-checkpoint.ipynb b/.ipynb_checkpoints/raremodel-nb-checkpoint.ipynb index ce13a76..e2d3820 100644 --- a/.ipynb_checkpoints/raremodel-nb-checkpoint.ipynb +++ b/.ipynb_checkpoints/raremodel-nb-checkpoint.ipynb @@ -65,7 +65,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 2, "metadata": {}, "outputs": [], "source": [ @@ -269,7 +269,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 3, "metadata": {}, "outputs": [], "source": [ @@ -304,7 +304,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 4, "metadata": {}, "outputs": [], "source": [ @@ -348,7 +348,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 5, "metadata": {}, "outputs": [], "source": [ @@ -376,9 +376,19 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 6, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "WARNING:tensorflow:From c:\\users\\sa_li\\.conda\\envs\\rmd\\lib\\site-packages\\tensorflow\\python\\ops\\resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n", + "Instructions for updating:\n", + "Colocations handled automatically by placer.\n" + ] + } + ], "source": [ "#jpsi\n", "\n", @@ -387,8 +397,8 @@ "\n", "jpsi_m = zfit.Parameter(\"jpsi_m\", ztf.constant(jpsi_mass), floating = False)\n", "jpsi_w = zfit.Parameter(\"jpsi_w\", ztf.constant(jpsi_width), floating = False)\n", - "jpsi_p = zfit.Parameter(\"jpsi_p\", ztf.constant(jpsi_phase), floating = False)\n", - "jpsi_s = zfit.Parameter(\"jpsi_s\", ztf.constant(jpsi_scale), floating = False)\n", + "jpsi_p = zfit.Parameter(\"jpsi_p\", ztf.constant(jpsi_phase))\n", + "jpsi_s = zfit.Parameter(\"jpsi_s\", ztf.constant(jpsi_scale))\n", "\n", "#psi2s\n", "\n", @@ -397,17 +407,17 @@ "\n", "psi2s_m = zfit.Parameter(\"psi2s_m\", ztf.constant(psi2s_mass), floating = False)\n", "psi2s_w = zfit.Parameter(\"psi2s_w\", ztf.constant(psi2s_width), floating = False)\n", - "psi2s_p = zfit.Parameter(\"psi2s_p\", ztf.constant(psi2s_phase), floating = False)\n", - "psi2s_s = zfit.Parameter(\"psi2s_s\", ztf.constant(psi2s_scale), floating = False)\n", + "psi2s_p = zfit.Parameter(\"psi2s_p\", ztf.constant(psi2s_phase))\n", + "psi2s_s = zfit.Parameter(\"psi2s_s\", ztf.constant(psi2s_scale))\n", "\n", "#cusp\n", "\n", "cusp_mass, sigma_R, sigma_L, cusp_scale = 3550, 3e-7, 200, 0\n", "\n", - "cusp_m = zfit.Parameter(\"cusp_m\", ztf.constant(cusp_mass))\n", - "sig_L = zfit.Parameter(\"sig_L\", ztf.constant(sigma_L))\n", - "sig_R = zfit.Parameter(\"sig_R\", ztf.constant(sigma_R))\n", - "cusp_s = zfit.Parameter(\"cusp_s\", ztf.constant(cusp_scale))" + "cusp_m = zfit.Parameter(\"cusp_m\", ztf.constant(cusp_mass), floating = False)\n", + "sig_L = zfit.Parameter(\"sig_L\", ztf.constant(sigma_L), floating = False)\n", + "sig_R = zfit.Parameter(\"sig_R\", ztf.constant(sigma_R), floating = False)\n", + "cusp_s = zfit.Parameter(\"cusp_s\", ztf.constant(cusp_scale), floating = False)" ] }, { @@ -419,7 +429,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 7, "metadata": {}, "outputs": [], "source": [ @@ -439,7 +449,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 8, "metadata": {}, "outputs": [], "source": [ @@ -479,9 +489,29 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 25, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "0.09\n" + ] + }, + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], "source": [ "plt.clf()\n", "# plt.plot(x_part, calcs, '.')\n", @@ -504,7 +534,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 10, "metadata": {}, "outputs": [], "source": [ @@ -516,7 +546,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 11, "metadata": {}, "outputs": [], "source": [ @@ -528,7 +558,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 12, "metadata": {}, "outputs": [], "source": [ @@ -560,16 +590,24 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 13, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "5404695.652173913\n" + ] + } + ], "source": [ "print(36000*(1+ pdg[\"jpsi_BR\"]/pdg[\"NR_BR\"] + pdg[\"psi2s_BR\"]/pdg[\"NR_BR\"]))" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 14, "metadata": {}, "outputs": [], "source": [ @@ -590,7 +628,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 15, "metadata": {}, "outputs": [], "source": [ @@ -599,43 +637,48 @@ "\n", "calls = int(nevents/event_stack + 1)\n", "\n", - "total_samp = []\n", + "# total_samp = []\n", "\n", - "start = time.time()\n", + "# start = time.time()\n", "\n", - "for call in range(calls):\n", - " samp = total_f.sample(n=event_stack)\n", - " sam = samp.unstack_x()\n", - " sam = zfit.run(sam)\n", - " clear_output(wait=True)\n", + "# for call in range(calls):\n", + "# samp = total_f.sample(n=event_stack)\n", + "# sam = samp.unstack_x()\n", + "# sam = zfit.run(sam)\n", + "# clear_output(wait=True)\n", " \n", - " print(\"{0}/{1}\".format(call + 1, calls))\n", - " print(\"Time taken: {}\".format(display_time(int(time.time() - start))))\n", - " c = call + 1\n", - " print(\"Projected time left: {}\".format(display_time(int((time.time() - start)/c*(calls-c)))))\n", + "# print(\"{0}/{1}\".format(call + 1, calls))\n", + "# print(\"Time taken: {}\".format(display_time(int(time.time() - start))))\n", + "# c = call + 1\n", + "# print(\"Projected time left: {}\".format(display_time(int((time.time() - start)/c*(calls-c)))))\n", " \n", " \n", - " with open(\"data/zfit_toys/toy_1/{}.pkl\".format(call), \"wb\") as f:\n", - " pkl.dump(sam, f, pkl.HIGHEST_PROTOCOL)" + "# with open(\"data/zfit_toys/toy_1/{}.pkl\".format(call), \"wb\") as f:\n", + "# pkl.dump(sam, f, pkl.HIGHEST_PROTOCOL)" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 16, "metadata": {}, "outputs": [], "source": [ - "# with open(\"data/zfit_toys/test_toy.pkl\", \"wb\") as f:\n", - "# pkl.dump(total_samp, f, pkl.HIGHEST_PROTOCOL)\n", - " \n", - "print(\"Time to generate full toy: {} s\".format(int(time.time()-start)))" + "# print(\"Time to generate full toy: {} s\".format(int(time.time()-start)))" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 17, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(5404696,)\n" + ] + } + ], "source": [ "total_samp = []\n", "\n", @@ -644,25 +687,42 @@ " sam = pkl.load(input_file)\n", " total_samp = np.append(total_samp, sam)\n", "\n", + "total_samp = total_samp.astype('float64')\n", + "\n", + "data2 = zfit.data.Data.from_numpy(array=total_samp[:int(nevents)], obs=obs)\n", + "\n", "print(total_samp[:nevents].shape)" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 18, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], "source": [ "bins = int((x_max-x_min)/7)\n", "\n", - "calcs = zfit.run(total_test_tf(samp))\n", + "# calcs = zfit.run(total_test_tf(samp))\n", "\n", "plt.hist(total_samp, bins = bins, range = (x_min,x_max))\n", "\n", "# plt.plot(sam, calcs, '.')\n", "# plt.plot(test_q, calcs_test)\n", - "# plt.ylim(0, 200)\n", - "# plt.xlim(3000, 3750)\n", + "plt.ylim(6000, 10000)\n", + "plt.xlim(3000, 3750)\n", "\n", "plt.savefig('test.png')" ] @@ -676,7 +736,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 19, "metadata": {}, "outputs": [], "source": [ @@ -706,22 +766,349 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 20, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
FCN = 19224270.58100143TOTAL NCALL = 58NCALLS = 58
EDM = 1.523018659051133e-07GOAL EDM = 5e-06\n", + " UP = 0.5
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
ValidValid ParamAccurate CovarPosDefMade PosDef
TrueTrueTrueTrueFalse
Hesse FailHasCovAbove EDMReach calllim
FalseTrueFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
+NameValueHesse ErrorMinos Error-Minos Error+Limit-Limit+Fixed?
0psi2s_p-9.424780.00651368No
1psi2s_s74.73490.048852No
2jpsi_p-9.424780.00631191No
3jpsi_s444.4950.220516No
\n", + "
\n",
+       "\n",
+       "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for psi2s_p: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.0065151408893557810.006517797883262716
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for psi2s_s: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.049227653881089160.0492610170079898
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for jpsi_p: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.006311001681890770.006310920173057535
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for jpsi_s: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.220659289686720880.22053487734368604
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "psi2s_p: ^{+0.006517797883262716}_{-0.006515140889355781}\n", + "psi2s_s: ^{+0.0492610170079898}_{-0.04922765388108916}\n", + "jpsi_p: ^{+0.006310920173057535}_{-0.00631100168189077}\n", + "jpsi_s: ^{+0.22053487734368604}_{-0.22065928968672088}\n", + "Function minimum: 19224270.58100143\n" + ] + } + ], "source": [ - "# nll = zfit.loss.UnbinnedNLL(model=total_f, data=data, fit_range = (x_min, x_max))\n", + "nll = zfit.loss.UnbinnedNLL(model=total_f, data=data2, fit_range = (x_min, x_max))\n", "\n", - "# minimizer = zfit.minimize.MinuitMinimizer()\n", + "minimizer = zfit.minimize.MinuitMinimizer()\n", "# minimizer._use_tfgrad = False\n", - "# result = minimizer.minimize(nll)\n", + "result = minimizer.minimize(nll)\n", "\n", - "# param_errors = result.error()\n", + "param_errors = result.error()\n", "\n", - "# for var, errors in param_errors.items():\n", - "# print('{}: ^{{+{}}}_{{{}}}'.format(var.name, errors['upper'], errors['lower']))\n", + "for var, errors in param_errors.items():\n", + " print('{}: ^{{+{}}}_{{{}}}'.format(var.name, errors['upper'], errors['lower']))\n", "\n", - "# print(\"Function minimum:\", result.fmin)" + "print(\"Function minimum:\", result.fmin)" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "-3.1420346928204133" + ] + }, + "execution_count": 21, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "(-9.42522+2*np.pi)#/np.pi" ] }, { @@ -729,41 +1116,11 @@ "execution_count": null, "metadata": {}, "outputs": [], - "source": [ - "# samp = total_f.sample(n=nevents)" - ] + "source": [] }, { "cell_type": "code", - "execution_count": null, - "metadata": {}, - "outputs": [], - "source": [ - "# sam = samp.unstack_x()\n", - "\n", - "# sam = zfit.run(sam)\n", - "\n", - "# bins = int((x_max-x_min)/7)\n", - "\n", - "# calcs = zfit.run(total_test_tf(samp))\n", - "\n", - "# plt.clf()\n", - "\n", - "# plt.hist(sam, bins = bins, range = (x_min,x_max))\n", - "\n", - "# # plt.plot(sam, calcs, '.')\n", - "# # plt.plot(test_q, calcs_test)\n", - "# # plt.ylim(0, 0.0000007)\n", - "# # plt.xlim(3000, 3750)\n", - "\n", - "# plt.ylim(0,1000)\n", - "\n", - "# plt.savefig('test.png')" - ] - }, - { - "cell_type": "code", - "execution_count": 2, + "execution_count": 22, "metadata": {}, "outputs": [ { @@ -772,7 +1129,7 @@ "'5 h, 55 min'" ] }, - "execution_count": 2, + "execution_count": 22, "metadata": {}, "output_type": "execute_result" } @@ -783,10 +1140,20 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 23, "metadata": {}, - "outputs": [], - "source": [] + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "6 h, 12 min\n" + ] + } + ], + "source": [ + "print(display_time(22376))" + ] }, { "cell_type": "code", diff --git a/data/zfit_toys/toy_1/0.pkl b/data/zfit_toys/toy_1/0.pkl index a5a5584..e3599d9 100644 --- a/data/zfit_toys/toy_1/0.pkl +++ b/data/zfit_toys/toy_1/0.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/1.pkl b/data/zfit_toys/toy_1/1.pkl index 0fe3188..319eb8b 100644 --- a/data/zfit_toys/toy_1/1.pkl +++ b/data/zfit_toys/toy_1/1.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/10.pkl b/data/zfit_toys/toy_1/10.pkl index 32c251f..423bda4 100644 --- a/data/zfit_toys/toy_1/10.pkl +++ b/data/zfit_toys/toy_1/10.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/11.pkl b/data/zfit_toys/toy_1/11.pkl new file mode 100644 index 0000000..117c716 --- /dev/null +++ b/data/zfit_toys/toy_1/11.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/12.pkl b/data/zfit_toys/toy_1/12.pkl new file mode 100644 index 0000000..75f43d2 --- /dev/null +++ b/data/zfit_toys/toy_1/12.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/13.pkl b/data/zfit_toys/toy_1/13.pkl new file mode 100644 index 0000000..2f53882 --- /dev/null +++ b/data/zfit_toys/toy_1/13.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/14.pkl b/data/zfit_toys/toy_1/14.pkl new file mode 100644 index 0000000..3b07370 --- /dev/null +++ b/data/zfit_toys/toy_1/14.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/15.pkl b/data/zfit_toys/toy_1/15.pkl new file mode 100644 index 0000000..a465ae6 --- /dev/null +++ b/data/zfit_toys/toy_1/15.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/16.pkl b/data/zfit_toys/toy_1/16.pkl new file mode 100644 index 0000000..373b614 --- /dev/null +++ b/data/zfit_toys/toy_1/16.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/17.pkl b/data/zfit_toys/toy_1/17.pkl new file mode 100644 index 0000000..ad9e2a2 --- /dev/null +++ b/data/zfit_toys/toy_1/17.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/18.pkl b/data/zfit_toys/toy_1/18.pkl new file mode 100644 index 0000000..f65a61f --- /dev/null +++ b/data/zfit_toys/toy_1/18.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/19.pkl b/data/zfit_toys/toy_1/19.pkl new file mode 100644 index 0000000..6aa8f65 --- /dev/null +++ b/data/zfit_toys/toy_1/19.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/2.pkl b/data/zfit_toys/toy_1/2.pkl index 75587e5..dbd1dde 100644 --- a/data/zfit_toys/toy_1/2.pkl +++ b/data/zfit_toys/toy_1/2.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/20.pkl b/data/zfit_toys/toy_1/20.pkl new file mode 100644 index 0000000..568f24a --- /dev/null +++ b/data/zfit_toys/toy_1/20.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/21.pkl b/data/zfit_toys/toy_1/21.pkl new file mode 100644 index 0000000..dd8e221 --- /dev/null +++ b/data/zfit_toys/toy_1/21.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/22.pkl b/data/zfit_toys/toy_1/22.pkl new file mode 100644 index 0000000..d5de54c --- /dev/null +++ b/data/zfit_toys/toy_1/22.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/23.pkl b/data/zfit_toys/toy_1/23.pkl new file mode 100644 index 0000000..0c41e46 --- /dev/null +++ b/data/zfit_toys/toy_1/23.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/24.pkl b/data/zfit_toys/toy_1/24.pkl new file mode 100644 index 0000000..89df721 --- /dev/null +++ b/data/zfit_toys/toy_1/24.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/25.pkl b/data/zfit_toys/toy_1/25.pkl new file mode 100644 index 0000000..15ee33d --- /dev/null +++ b/data/zfit_toys/toy_1/25.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/26.pkl b/data/zfit_toys/toy_1/26.pkl new file mode 100644 index 0000000..9dbb293 --- /dev/null +++ b/data/zfit_toys/toy_1/26.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/27.pkl b/data/zfit_toys/toy_1/27.pkl new file mode 100644 index 0000000..e68208f --- /dev/null +++ b/data/zfit_toys/toy_1/27.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/28.pkl b/data/zfit_toys/toy_1/28.pkl new file mode 100644 index 0000000..fa10112 --- /dev/null +++ b/data/zfit_toys/toy_1/28.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/29.pkl b/data/zfit_toys/toy_1/29.pkl new file mode 100644 index 0000000..c6adfb9 --- /dev/null +++ b/data/zfit_toys/toy_1/29.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/3.pkl b/data/zfit_toys/toy_1/3.pkl index 0ee2612..4af34bf 100644 --- a/data/zfit_toys/toy_1/3.pkl +++ b/data/zfit_toys/toy_1/3.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/30.pkl b/data/zfit_toys/toy_1/30.pkl new file mode 100644 index 0000000..9a781ad --- /dev/null +++ b/data/zfit_toys/toy_1/30.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/31.pkl b/data/zfit_toys/toy_1/31.pkl new file mode 100644 index 0000000..4a64c1d --- /dev/null +++ b/data/zfit_toys/toy_1/31.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/32.pkl b/data/zfit_toys/toy_1/32.pkl new file mode 100644 index 0000000..e6f4b01 --- /dev/null +++ b/data/zfit_toys/toy_1/32.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/33.pkl b/data/zfit_toys/toy_1/33.pkl new file mode 100644 index 0000000..7bceb8f --- /dev/null +++ b/data/zfit_toys/toy_1/33.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/34.pkl b/data/zfit_toys/toy_1/34.pkl new file mode 100644 index 0000000..2c772ce --- /dev/null +++ b/data/zfit_toys/toy_1/34.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/35.pkl b/data/zfit_toys/toy_1/35.pkl new file mode 100644 index 0000000..141fabb --- /dev/null +++ b/data/zfit_toys/toy_1/35.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/36.pkl b/data/zfit_toys/toy_1/36.pkl new file mode 100644 index 0000000..ccf5be8 --- /dev/null +++ b/data/zfit_toys/toy_1/36.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/37.pkl b/data/zfit_toys/toy_1/37.pkl new file mode 100644 index 0000000..2456338 --- /dev/null +++ b/data/zfit_toys/toy_1/37.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/38.pkl b/data/zfit_toys/toy_1/38.pkl new file mode 100644 index 0000000..dac0815 --- /dev/null +++ b/data/zfit_toys/toy_1/38.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/39.pkl b/data/zfit_toys/toy_1/39.pkl new file mode 100644 index 0000000..5ac71b6 --- /dev/null +++ b/data/zfit_toys/toy_1/39.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/4.pkl b/data/zfit_toys/toy_1/4.pkl index 017deb2..391e9b5 100644 --- a/data/zfit_toys/toy_1/4.pkl +++ b/data/zfit_toys/toy_1/4.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/40.pkl b/data/zfit_toys/toy_1/40.pkl new file mode 100644 index 0000000..0c18293 --- /dev/null +++ b/data/zfit_toys/toy_1/40.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/41.pkl b/data/zfit_toys/toy_1/41.pkl new file mode 100644 index 0000000..c13564a --- /dev/null +++ b/data/zfit_toys/toy_1/41.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/42.pkl b/data/zfit_toys/toy_1/42.pkl new file mode 100644 index 0000000..84eafa1 --- /dev/null +++ b/data/zfit_toys/toy_1/42.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/43.pkl b/data/zfit_toys/toy_1/43.pkl new file mode 100644 index 0000000..37c50b8 --- /dev/null +++ b/data/zfit_toys/toy_1/43.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/44.pkl b/data/zfit_toys/toy_1/44.pkl new file mode 100644 index 0000000..32433f5 --- /dev/null +++ b/data/zfit_toys/toy_1/44.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/45.pkl b/data/zfit_toys/toy_1/45.pkl new file mode 100644 index 0000000..1fba254 --- /dev/null +++ b/data/zfit_toys/toy_1/45.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/46.pkl b/data/zfit_toys/toy_1/46.pkl new file mode 100644 index 0000000..5708771 --- /dev/null +++ b/data/zfit_toys/toy_1/46.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/47.pkl b/data/zfit_toys/toy_1/47.pkl new file mode 100644 index 0000000..5758900 --- /dev/null +++ b/data/zfit_toys/toy_1/47.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/48.pkl b/data/zfit_toys/toy_1/48.pkl new file mode 100644 index 0000000..9b2a2b7 --- /dev/null +++ b/data/zfit_toys/toy_1/48.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/49.pkl b/data/zfit_toys/toy_1/49.pkl new file mode 100644 index 0000000..720e13d --- /dev/null +++ b/data/zfit_toys/toy_1/49.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/5.pkl b/data/zfit_toys/toy_1/5.pkl index 3197c30..ed572a1 100644 --- a/data/zfit_toys/toy_1/5.pkl +++ b/data/zfit_toys/toy_1/5.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/50.pkl b/data/zfit_toys/toy_1/50.pkl new file mode 100644 index 0000000..24e4f03 --- /dev/null +++ b/data/zfit_toys/toy_1/50.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/51.pkl b/data/zfit_toys/toy_1/51.pkl new file mode 100644 index 0000000..2bf7b34 --- /dev/null +++ b/data/zfit_toys/toy_1/51.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/52.pkl b/data/zfit_toys/toy_1/52.pkl new file mode 100644 index 0000000..2e3ff8c --- /dev/null +++ b/data/zfit_toys/toy_1/52.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/53.pkl b/data/zfit_toys/toy_1/53.pkl new file mode 100644 index 0000000..62a3d60 --- /dev/null +++ b/data/zfit_toys/toy_1/53.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/54.pkl b/data/zfit_toys/toy_1/54.pkl new file mode 100644 index 0000000..4613047 --- /dev/null +++ b/data/zfit_toys/toy_1/54.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/6.pkl b/data/zfit_toys/toy_1/6.pkl index 269a4cc..c2853d7 100644 --- a/data/zfit_toys/toy_1/6.pkl +++ b/data/zfit_toys/toy_1/6.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/7.pkl b/data/zfit_toys/toy_1/7.pkl index 27c52f1..bf66a13 100644 --- a/data/zfit_toys/toy_1/7.pkl +++ b/data/zfit_toys/toy_1/7.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/8.pkl b/data/zfit_toys/toy_1/8.pkl index 68079f2..1bb89f1 100644 --- a/data/zfit_toys/toy_1/8.pkl +++ b/data/zfit_toys/toy_1/8.pkl Binary files differ diff --git a/data/zfit_toys/toy_1/9.pkl b/data/zfit_toys/toy_1/9.pkl index b3aedb0..92c2d22 100644 --- a/data/zfit_toys/toy_1/9.pkl +++ b/data/zfit_toys/toy_1/9.pkl Binary files differ diff --git a/raremodel-nb.ipynb b/raremodel-nb.ipynb index ce13a76..e2d3820 100644 --- a/raremodel-nb.ipynb +++ b/raremodel-nb.ipynb @@ -65,7 +65,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 2, "metadata": {}, "outputs": [], "source": [ @@ -269,7 +269,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 3, "metadata": {}, "outputs": [], "source": [ @@ -304,7 +304,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 4, "metadata": {}, "outputs": [], "source": [ @@ -348,7 +348,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 5, "metadata": {}, "outputs": [], "source": [ @@ -376,9 +376,19 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 6, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "WARNING:tensorflow:From c:\\users\\sa_li\\.conda\\envs\\rmd\\lib\\site-packages\\tensorflow\\python\\ops\\resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n", + "Instructions for updating:\n", + "Colocations handled automatically by placer.\n" + ] + } + ], "source": [ "#jpsi\n", "\n", @@ -387,8 +397,8 @@ "\n", "jpsi_m = zfit.Parameter(\"jpsi_m\", ztf.constant(jpsi_mass), floating = False)\n", "jpsi_w = zfit.Parameter(\"jpsi_w\", ztf.constant(jpsi_width), floating = False)\n", - "jpsi_p = zfit.Parameter(\"jpsi_p\", ztf.constant(jpsi_phase), floating = False)\n", - "jpsi_s = zfit.Parameter(\"jpsi_s\", ztf.constant(jpsi_scale), floating = False)\n", + "jpsi_p = zfit.Parameter(\"jpsi_p\", ztf.constant(jpsi_phase))\n", + "jpsi_s = zfit.Parameter(\"jpsi_s\", ztf.constant(jpsi_scale))\n", "\n", "#psi2s\n", "\n", @@ -397,17 +407,17 @@ "\n", "psi2s_m = zfit.Parameter(\"psi2s_m\", ztf.constant(psi2s_mass), floating = False)\n", "psi2s_w = zfit.Parameter(\"psi2s_w\", ztf.constant(psi2s_width), floating = False)\n", - "psi2s_p = zfit.Parameter(\"psi2s_p\", ztf.constant(psi2s_phase), floating = False)\n", - "psi2s_s = zfit.Parameter(\"psi2s_s\", ztf.constant(psi2s_scale), floating = False)\n", + "psi2s_p = zfit.Parameter(\"psi2s_p\", ztf.constant(psi2s_phase))\n", + "psi2s_s = zfit.Parameter(\"psi2s_s\", ztf.constant(psi2s_scale))\n", "\n", "#cusp\n", "\n", "cusp_mass, sigma_R, sigma_L, cusp_scale = 3550, 3e-7, 200, 0\n", "\n", - "cusp_m = zfit.Parameter(\"cusp_m\", ztf.constant(cusp_mass))\n", - "sig_L = zfit.Parameter(\"sig_L\", ztf.constant(sigma_L))\n", - "sig_R = zfit.Parameter(\"sig_R\", ztf.constant(sigma_R))\n", - "cusp_s = zfit.Parameter(\"cusp_s\", ztf.constant(cusp_scale))" + "cusp_m = zfit.Parameter(\"cusp_m\", ztf.constant(cusp_mass), floating = False)\n", + "sig_L = zfit.Parameter(\"sig_L\", ztf.constant(sigma_L), floating = False)\n", + "sig_R = zfit.Parameter(\"sig_R\", ztf.constant(sigma_R), floating = False)\n", + "cusp_s = zfit.Parameter(\"cusp_s\", ztf.constant(cusp_scale), floating = False)" ] }, { @@ -419,7 +429,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 7, "metadata": {}, "outputs": [], "source": [ @@ -439,7 +449,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 8, "metadata": {}, "outputs": [], "source": [ @@ -479,9 +489,29 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 25, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "0.09\n" + ] + }, + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], "source": [ "plt.clf()\n", "# plt.plot(x_part, calcs, '.')\n", @@ -504,7 +534,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 10, "metadata": {}, "outputs": [], "source": [ @@ -516,7 +546,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 11, "metadata": {}, "outputs": [], "source": [ @@ -528,7 +558,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 12, "metadata": {}, "outputs": [], "source": [ @@ -560,16 +590,24 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 13, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "5404695.652173913\n" + ] + } + ], "source": [ "print(36000*(1+ pdg[\"jpsi_BR\"]/pdg[\"NR_BR\"] + pdg[\"psi2s_BR\"]/pdg[\"NR_BR\"]))" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 14, "metadata": {}, "outputs": [], "source": [ @@ -590,7 +628,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 15, "metadata": {}, "outputs": [], "source": [ @@ -599,43 +637,48 @@ "\n", "calls = int(nevents/event_stack + 1)\n", "\n", - "total_samp = []\n", + "# total_samp = []\n", "\n", - "start = time.time()\n", + "# start = time.time()\n", "\n", - "for call in range(calls):\n", - " samp = total_f.sample(n=event_stack)\n", - " sam = samp.unstack_x()\n", - " sam = zfit.run(sam)\n", - " clear_output(wait=True)\n", + "# for call in range(calls):\n", + "# samp = total_f.sample(n=event_stack)\n", + "# sam = samp.unstack_x()\n", + "# sam = zfit.run(sam)\n", + "# clear_output(wait=True)\n", " \n", - " print(\"{0}/{1}\".format(call + 1, calls))\n", - " print(\"Time taken: {}\".format(display_time(int(time.time() - start))))\n", - " c = call + 1\n", - " print(\"Projected time left: {}\".format(display_time(int((time.time() - start)/c*(calls-c)))))\n", + "# print(\"{0}/{1}\".format(call + 1, calls))\n", + "# print(\"Time taken: {}\".format(display_time(int(time.time() - start))))\n", + "# c = call + 1\n", + "# print(\"Projected time left: {}\".format(display_time(int((time.time() - start)/c*(calls-c)))))\n", " \n", " \n", - " with open(\"data/zfit_toys/toy_1/{}.pkl\".format(call), \"wb\") as f:\n", - " pkl.dump(sam, f, pkl.HIGHEST_PROTOCOL)" + "# with open(\"data/zfit_toys/toy_1/{}.pkl\".format(call), \"wb\") as f:\n", + "# pkl.dump(sam, f, pkl.HIGHEST_PROTOCOL)" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 16, "metadata": {}, "outputs": [], "source": [ - "# with open(\"data/zfit_toys/test_toy.pkl\", \"wb\") as f:\n", - "# pkl.dump(total_samp, f, pkl.HIGHEST_PROTOCOL)\n", - " \n", - "print(\"Time to generate full toy: {} s\".format(int(time.time()-start)))" + "# print(\"Time to generate full toy: {} s\".format(int(time.time()-start)))" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 17, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "(5404696,)\n" + ] + } + ], "source": [ "total_samp = []\n", "\n", @@ -644,25 +687,42 @@ " sam = pkl.load(input_file)\n", " total_samp = np.append(total_samp, sam)\n", "\n", + "total_samp = total_samp.astype('float64')\n", + "\n", + "data2 = zfit.data.Data.from_numpy(array=total_samp[:int(nevents)], obs=obs)\n", + "\n", "print(total_samp[:nevents].shape)" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 18, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "image/png": "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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "needs_background": "light" + }, + "output_type": "display_data" + } + ], "source": [ "bins = int((x_max-x_min)/7)\n", "\n", - "calcs = zfit.run(total_test_tf(samp))\n", + "# calcs = zfit.run(total_test_tf(samp))\n", "\n", "plt.hist(total_samp, bins = bins, range = (x_min,x_max))\n", "\n", "# plt.plot(sam, calcs, '.')\n", "# plt.plot(test_q, calcs_test)\n", - "# plt.ylim(0, 200)\n", - "# plt.xlim(3000, 3750)\n", + "plt.ylim(6000, 10000)\n", + "plt.xlim(3000, 3750)\n", "\n", "plt.savefig('test.png')" ] @@ -676,7 +736,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 19, "metadata": {}, "outputs": [], "source": [ @@ -706,22 +766,349 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 20, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
FCN = 19224270.58100143TOTAL NCALL = 58NCALLS = 58
EDM = 1.523018659051133e-07GOAL EDM = 5e-06\n", + " UP = 0.5
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
ValidValid ParamAccurate CovarPosDefMade PosDef
TrueTrueTrueTrueFalse
Hesse FailHasCovAbove EDMReach calllim
FalseTrueFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
+NameValueHesse ErrorMinos Error-Minos Error+Limit-Limit+Fixed?
0psi2s_p-9.424780.00651368No
1psi2s_s74.73490.048852No
2jpsi_p-9.424780.00631191No
3jpsi_s444.4950.220516No
\n", + "
\n",
+       "\n",
+       "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for psi2s_p: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.0065151408893557810.006517797883262716
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for psi2s_s: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.049227653881089160.0492610170079898
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for jpsi_p: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.006311001681890770.006310920173057535
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/html": [ + "Minos status for jpsi_s: VALID\n", + "\n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + " \n", + "
Error-0.220659289686720880.22053487734368604
ValidTrueTrue
At LimitFalseFalse
Max FCNFalseFalse
New MinFalseFalse
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "psi2s_p: ^{+0.006517797883262716}_{-0.006515140889355781}\n", + "psi2s_s: ^{+0.0492610170079898}_{-0.04922765388108916}\n", + "jpsi_p: ^{+0.006310920173057535}_{-0.00631100168189077}\n", + "jpsi_s: ^{+0.22053487734368604}_{-0.22065928968672088}\n", + "Function minimum: 19224270.58100143\n" + ] + } + ], "source": [ - "# nll = zfit.loss.UnbinnedNLL(model=total_f, data=data, fit_range = (x_min, x_max))\n", + "nll = zfit.loss.UnbinnedNLL(model=total_f, data=data2, fit_range = (x_min, x_max))\n", "\n", - "# minimizer = zfit.minimize.MinuitMinimizer()\n", + "minimizer = zfit.minimize.MinuitMinimizer()\n", "# minimizer._use_tfgrad = False\n", - "# result = minimizer.minimize(nll)\n", + "result = minimizer.minimize(nll)\n", "\n", - "# param_errors = result.error()\n", + "param_errors = result.error()\n", "\n", - "# for var, errors in param_errors.items():\n", - "# print('{}: ^{{+{}}}_{{{}}}'.format(var.name, errors['upper'], errors['lower']))\n", + "for var, errors in param_errors.items():\n", + " print('{}: ^{{+{}}}_{{{}}}'.format(var.name, errors['upper'], errors['lower']))\n", "\n", - "# print(\"Function minimum:\", result.fmin)" + "print(\"Function minimum:\", result.fmin)" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "-3.1420346928204133" + ] + }, + "execution_count": 21, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "(-9.42522+2*np.pi)#/np.pi" ] }, { @@ -729,41 +1116,11 @@ "execution_count": null, "metadata": {}, "outputs": [], - "source": [ - "# samp = total_f.sample(n=nevents)" - ] + "source": [] }, { "cell_type": "code", - "execution_count": null, - "metadata": {}, - "outputs": [], - "source": [ - "# sam = samp.unstack_x()\n", - "\n", - "# sam = zfit.run(sam)\n", - "\n", - "# bins = int((x_max-x_min)/7)\n", - "\n", - "# calcs = zfit.run(total_test_tf(samp))\n", - "\n", - "# plt.clf()\n", - "\n", - "# plt.hist(sam, bins = bins, range = (x_min,x_max))\n", - "\n", - "# # plt.plot(sam, calcs, '.')\n", - "# # plt.plot(test_q, calcs_test)\n", - "# # plt.ylim(0, 0.0000007)\n", - "# # plt.xlim(3000, 3750)\n", - "\n", - "# plt.ylim(0,1000)\n", - "\n", - "# plt.savefig('test.png')" - ] - }, - { - "cell_type": "code", - "execution_count": 2, + "execution_count": 22, "metadata": {}, "outputs": [ { @@ -772,7 +1129,7 @@ "'5 h, 55 min'" ] }, - "execution_count": 2, + "execution_count": 22, "metadata": {}, "output_type": "execute_result" } @@ -783,10 +1140,20 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 23, "metadata": {}, - "outputs": [], - "source": [] + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "6 h, 12 min\n" + ] + } + ], + "source": [ + "print(display_time(22376))" + ] }, { "cell_type": "code",