soedinglab · benjamin-lieser · Feb 18, 2026 · Dec 20, 2025 · Dec 20, 2025 · Feb 13, 2026
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -7,5 +7,5 @@ members = [
 ]
 
 [profile.release]
-lto = true
-debug = true
+lto = false
+debug = false
diff --git a/benchmark_test/test.py b/benchmark_test/test.py
@@ -34,29 +34,23 @@ def gen_data_single_model(t_dtype, dim, seed):
     sqrt_pi = sqrt_pi[0:1]
     return S, sqrt_pi
 
-def helper_test(dtype, dim : int, gradients: bool, single_model: bool = False, gpu: bool = False):
-    if dtype == "f32":
-        t_dtype = torch.float32
-        np_dtype = np.float32
-    else:
-        t_dtype = torch.float64
-        np_dtype = np.float64
+def helper_test(dim : int, gradients: bool, single_model: bool = False):
+    t_dtype = torch.float64
+    np_dtype = np.float64
 
-    if dtype == "f32":
-        rtol = 1e-2
-        atol = 1e-2
-    else:
-        rtol = 1e-4
-        atol = 1e-4
+    rtol = 1e-4
+    atol = 1e-4
 
     torch_tree, parent_list, branch_lengths, leaf_log_p = gen_tree(t_dtype, dim)
 
+    leaf_pl = torch.exp(leaf_log_p)
+
     if single_model:
         data = [gen_data_single_model(t_dtype, dim, seed) for seed in range(10)]
     else:
         data = [gen_data(t_dtype, dim, seed) for seed in range(10)]
 
-    rust_tree = phylo_grad.FelsensteinTree(parent_list, branch_lengths.astype(np_dtype), leaf_log_p.numpy(), 1e-4, gpu)
+    rust_tree = phylo_grad.FelsensteinTree(parent_list, branch_lengths.astype(np_dtype), leaf_pl.numpy(), 1e-4)
 
 
     for i in range(10):
@@ -65,10 +59,7 @@ def helper_test(dtype, dim : int, gradients: bool, single_model: bool = False, g
 
 
         result = rust_tree.calculate_gradients(S.numpy(), sqrt_pi.numpy())
-        if gpu == False:
-            likelihoods = rust_tree.calculate_log_likelihoods(S.numpy(), sqrt_pi.numpy()) 
-            assert(np.allclose(likelihoods, result['log_likelihood'], rtol=1e-5))
-
+
         assert(np.allclose(result['log_likelihood'], torch_logP.numpy(), rtol=rtol))
 
         if gradients:
@@ -93,31 +84,14 @@ def helper_test(dtype, dim : int, gradients: bool, single_model: bool = False, g
             assert(np.allclose(result['grad_s'], torch_S_grad, rtol=rtol, atol=atol))
 
 def test_likelihood():
-    helper_test("f32", 4, False)
-    helper_test("f32", 20, False)
-    helper_test("f64", 4, False)
-    helper_test("f64", 20, False)
-
-def test_gpu_likelihood():
-    helper_test("f64", 4, False, gpu = True)
-    helper_test("f64", 20, False, gpu= True)
-    helper_test("f32", 4, False, gpu= True)
-    helper_test("f32", 20, False, gpu= True)
+    helper_test(4, False, True)
+    helper_test(4, False, False)
+    helper_test(20, False, True)
+    helper_test(20, False, False)
 
 def test_grads():
-    helper_test("f32", 4, True)
-    helper_test("f32", 20, True)
-    helper_test("f64", 4, True)
-    helper_test("f64", 20, True)
-
-def test_grads_single_model():
-    helper_test("f64", 4, True, single_model=True)
-    helper_test("f64", 20, True, single_model=True)
-    helper_test("f32", 4, True, single_model=True)
-    helper_test("f32", 20, True, single_model=True)
-
-def test_gpu_grads():
-    helper_test("f64", 4, True, gpu = True)
-    helper_test("f64", 20, True, gpu= True)
-    helper_test("f32", 4, True, gpu= True)
-    helper_test("f32", 20, True, gpu= True)
+    helper_test(4, True, True)
+    helper_test(4, True, False)
+    helper_test(20, True, True)
+    helper_test(20, True, False)
+
diff --git a/benchmark_test/test_tree.newick b/benchmark_test/test_tree.newick
@@ -1 +1 @@
-((((Tip_21:0.6650676023435903,Tip_22:0.024885436124515284):0.2056669529961157,((Tip_37:0.0337728838700196,Tip_38:0.9484958108725318):0.7273031959389747,Tip_36:0.4958302887613361):0.33757637305798605):0.8499954308656537,(Tip_4:0.8141476837951772,Tip_5:0.7402798051460897):0.70741682436221):0.813355,((((((Tip_83:0.17643378492960254,Tip_84:0.7953558100843726):0.14557382727056703,(Tip_43:0.8104945919986606,Tip_44:0.31449340792858616):0.5734673065270548):0.5607844652217767,Tip_18:0.17900274566087643):0.5063065847624981,(((Tip_32:0.6491605888497303,Tip_33:0.3429582140411475):0.14180550845435472,Tip_28:0.08365819037239011):0.5365343800267653,Tip_24:0.8800154043806051):0.998762888840805):0.8304610217387793,(Tip_0:0.7039506434317381,(Tip_25:0.7331478876918618,(Tip_27:0.9441692227658312,((Tip_57:0.38896701363191544,Tip_58:0.13379392008599617):0.8754816391828532,(Tip_60:0.352209387501356,Tip_61:0.3956010586841662):0.4583533711457218):0.5460177635944152):0.6361110453676238):0.6602589686792034):0.027727204798649534):0.6704649990971151,(((Tip_3:0.21104851572654998,(((((((((Tip_73:0.9483105414473861,Tip_74:0.6626788692704468):0.7822936616109846,Tip_59:0.12249378702437955):0.08222970026113245,(Tip_96:0.4290800932938288,Tip_97:0.24924474555329348):0.9179999611656243):0.7548297895532528,(Tip_98:0.9167802235932272,Tip_99:0.7204509004585486):0.8287579648322002):0.05271172049884924,((Tip_81:0.5413926753801398,Tip_82:0.4118750056383439):0.9486624612083683,Tip_39:0.29899889747606284):0.055068995802558016):0.2501971861428551,((Tip_86:0.0380499276137973,Tip_87:0.4256548943232936):0.07123509097099835,((((Tip_55:0.5710466520679875,Tip_56:0.14775864025032526):0.32953437140242836,Tip_47:0.7864549772499915):0.4280637902774943,Tip_45:0.13776127824637646):0.16954936978362714,Tip_23:0.17176851427365566):0.18988101382133601):0.9138259583612118):0.09048036473551212,(((Tip_92:0.404429134512315,Tip_93:0.047246278915062306):0.06224988908172349,((Tip_75:0.6566656102220733,Tip_76:0.6381056061984103):0.29787087139597435,Tip_40:0.5279535103620298):0.8286398644888646):0.19446009582787896,(Tip_34:0.26670755857819356,(Tip_69:0.10832233286814032,Tip_70:0.9365264322108605):0.41936902078708815):0.27879362246002265):0.5567981824116087):0.2694506279899209,Tip_6:0.9312226738371098):0.9215301310000562,(Tip_10:0.4514212322743643,Tip_11:0.18865283598284918):0.8275734698074786):0.23518611049075663):0.992212745439618,((Tip_19:0.8295636279502516,Tip_20:0.34263641161658875):0.9114454035382507,((((Tip_51:0.9220107008370401,Tip_52:0.8176402629230148):0.006951888793413944,(((Tip_94:0.3251911236094175,Tip_95:0.9900649501189612):0.2558363577627372,Tip_85:0.5404121010340436):0.9010280119121413,Tip_29:0.7328316282669919):0.9489623456648526):0.669567713578707,Tip_26:0.8016779879698788):0.6729191308869212,Tip_17:0.29215200135456015):0.31179956452592034):0.2543242891578164):0.5584080530741793,(Tip_66:0.8947516564428961,Tip_67:0.5767430884730821):0.2489036977476457):0.9656347634788244):0.239977,((((Tip_2:0.9464313983922135,(Tip_88:0.41442648404621196,Tip_89:0.5101163097014396):0.5466571358298574):0.6334947392181386,((((((Tip_12:0.11158676296777406,Tip_13:0.7279909147393833):0.2242187756664422,(Tip_41:0.7690871167673865,Tip_42:0.7384233696004908):0.9385579338831107):0.5705174119616021,(Tip_7:0.7360096495183802,Tip_8:0.25431554405864565):0.5988161405333698):0.9687384235590413,((Tip_14:0.49387953934218143,(Tip_62:0.9806096658151946,Tip_63:0.04665946024200317):0.21648740028859653):0.12806441473128458,(((Tip_71:0.4820889062355187,Tip_72:0.669077938124829):0.6907792707402931,Tip_46:0.7915022800264976):0.3224651326307264,Tip_1:0.37807097934208633):0.08262352954341846):0.5415349680132456):0.5368870623401885,(((Tip_9:0.30112516594260985,(Tip_30:0.8010675030911405,Tip_31:0.2197855899426761):0.2979712085110747):0.20561375454872075,(Tip_79:0.322022616870698,Tip_80:0.8405404258659477):0.3736763838229683):0.15489065294790658,((Tip_77:0.1631714880198354,Tip_78:0.14347186085240984):0.6350189977427474,((((Tip_90:0.8211951990555387,Tip_91:0.30263563690395334):0.6502375388726693,Tip_68:0.7812120071752737):0.36830312938864523,Tip_48:0.7430015399338462):0.17280797268648893,Tip_35:0.4635593017282353):0.7149095458288683):0.5852674316858171):0.8996082718667177):0.17732798167953295,((Tip_49:0.33471854305440063,Tip_50:0.11184498641576858):0.8048081839696574,(Tip_53:0.6764255828447392,Tip_54:0.11645530351988755):0.47997626175532904):0.7911201160390272):0.806790252892112):0.43691677945398477,(Tip_64:0.6545202475677102,Tip_65:0.8679242823464907):0.6299303286521293):0.7947232572644647,(Tip_15:0.7247905356618102,Tip_16:0.7397524628980965):0.10514061923847424):1.13067);
+(((Tip_3:0.10805328350042452,Tip_4:0.3315005520372256):0.7775108391547321,(((Tip_10:0.9543578556535617,Tip_11:0.46520848427708783):0.19055443061683935,Tip_5:0.8608787164185184):0.2282830452051495,(Tip_8:0.45877603959910257,Tip_9:0.2188740043797414):0.977427116635245):0.8761171479386695):1.02879,((((Tip_14:0.3495434398130265,Tip_15:0.7072552650535753):0.44843039151645087,(Tip_6:0.8903069753526421,Tip_7:0.11856013796286433):0.6507572690541433):0.7600580845054129,Tip_1:0.2230762517502976):0.04289178403051824,Tip_0:0.7254337418961002):0.143964,((Tip_12:0.535570065080058,Tip_13:0.9288624624322107):0.771915926537751,Tip_2:0.9590719826516791):0.939461);
diff --git a/gtr_optimize/Cargo.toml b/gtr_optimize/Cargo.toml
@@ -15,6 +15,7 @@ phylotree = "0.1.3"
 lazy_static = "1.5.0"
 gosh-lbfgs = "0.1.0"
 seq_io = "0.3.4"
-nalgebra = "0.33.2"
+nalgebra = "0.34.1"
 num-traits = {version="0.2.19", default-features=false, features = ["libm"]}
 rayon = "1.11.0"
+logsumexp = "0.1.0"
diff --git a/gtr_optimize/src/lib.rs b/gtr_optimize/src/lib.rs
@@ -4,11 +4,12 @@
 //! It has a global matrix mode a per column matrix mode, caled "local" here.
 
 use lazy_static::lazy_static;
+use logsumexp::LogSumExp;
 use nalgebra as na;
 use rayon::iter::{IntoParallelRefMutIterator, ParallelIterator};
 use std::collections::HashMap;
 
-use phylo_grad::{FelsensteinTree, FloatTrait};
+use phylo_grad::FelsensteinTree;
 
 lazy_static! {
     static ref AMINO_MAPPING: HashMap<u8, u8> = {
@@ -39,16 +40,15 @@ lazy_static! {
     };
 }
 
-pub fn seq2pll<F: FloatTrait>(seq: impl Iterator<Item = u8>) -> Vec<na::SVector<F, 20>> {
+pub fn seq2pll(seq: impl Iterator<Item = u8>) -> Vec<na::SVector<f64, 20>> {
     seq.map(|c| *AMINO_MAPPING.get(&c).unwrap_or(&20))
         .map(|idx| {
-            let mut v =
-                na::SVector::<F, 20>::from_element(<F as FloatTrait>::from_f64(f64::NEG_INFINITY));
+            let mut v = na::SVector::<f64, 20>::from_element(0.0);
             if idx < 20 {
-                v[idx as usize] = F::zero();
+                v[idx as usize] = 1.0;
             } else {
                 for i in 0..20 {
-                    v[i] = F::zero();
+                    v[i] = 1.0;
                 }
             }
             v
@@ -60,7 +60,7 @@ pub fn process_newick_alignment(
     newick: &str,
     sequences: &HashMap<String, Vec<u8>>,
 ) -> (
-    phylo_grad::FelsensteinTree<f64, 20>,
+    phylo_grad::FelsensteinTree<20>,
     Vec<Vec<na::SVector<f64, 20>>>,
 ) {
     let tree = phylotree::tree::Tree::from_newick(newick).unwrap();
@@ -112,12 +112,23 @@ pub fn process_newick_alignment(
                 column_seq[new_idx as usize] = seq[i];
             }
         }
-        leaf_pll.push(seq2pll::<f64>(column_seq.into_iter()));
+        leaf_pll.push(seq2pll(column_seq.into_iter()));
     }
 
-    let felsenstein = FelsensteinTree::<f64, 20>::new(&parents, &distances);
+    let felsenstein = FelsensteinTree::<20>::new(&parents, &distances);
     (felsenstein, leaf_pll)
 }
+/// Numerical stable softmax
+pub fn softmax<const N: usize>(x: &na::SVector<f64, N>) -> na::SVector<f64, N> {
+    let x_max = x.max();
+
+    let result = x.add_scalar(-x_max);
+
+    let mut result = result.map(|x| num_traits::Float::exp(x));
+
+    result /= result.sum();
+    result
+}
 
 pub fn optimize_gtr_local(newick: &str, sequences: &HashMap<String, Vec<u8>>) -> f64 {
     let (felsenstein, mut leaf_pll) = process_newick_alignment(newick, sequences);
@@ -135,7 +146,7 @@ pub fn optimize_gtr_local(newick: &str, sequences: &HashMap<String, Vec<u8>>) ->
 }
 
 fn optimize_gtr_single_side(
-    felsenstein: &FelsensteinTree<f64, 20>,
+    felsenstein: &FelsensteinTree<20>,
     log_pi_init: &[f64],
     log_p: &mut [na::SVector<f64, 20>],
 ) -> f64 {
@@ -198,7 +209,7 @@ fn optimize_gtr_single_side(
 
 pub fn optimize_gtr_global(newick: &str, sequences: &HashMap<String, Vec<u8>>) -> f64 {
     let (mut felsenstein, leaf_pll) = process_newick_alignment(newick, sequences);
-    felsenstein.bind_leaf_log_p(leaf_pll);
+    felsenstein.bind_leaf_pl(leaf_pll);
 
     let evaluate = |x: &[f64], g: &mut [f64]| {
         let log_R = &x[..190];
@@ -282,7 +293,7 @@ fn rate_matrix_backward(
     let d_logM = -d_logS.sum();
 
     let piRpi_max = data.piRpi.max();
-    let piRpi_exp: na::SMatrix<f64, 20, 20> = data.piRpi.map(|x| (x - piRpi_max).scalar_exp());
+    let piRpi_exp: na::SMatrix<f64, 20, 20> = data.piRpi.map(|x| (x - piRpi_max).exp());
     let piRpi_sum = piRpi_exp.sum();
 
     let d_piRpi = piRpi_exp.map(|x| x * d_logM / piRpi_sum);
@@ -304,7 +315,7 @@ fn rate_matrix_backward(
     let d_log_pi =
         d_log_pi + sqrt_pi_cotangent.component_mul(&data.log_pi.map(|x| 0.5 * (0.5 * x).exp()));
 
-    let softmax_log_pi_unorm = phylo_grad::softmax(&data.log_pi_unormalized);
+    let softmax_log_pi_unorm = softmax(&data.log_pi_unormalized);
 
     for i in 0..20 {
         grad_log_pi[i] = d_log_pi[i] - softmax_log_pi_unorm[i] * d_log_pi.sum();
@@ -316,7 +327,7 @@ fn rate_matrix_backward(
 fn rate_matrix(log_R: &[f64], log_pi_unormalized: &[f64]) -> RateMatrixData {
     let log_pi_unormalized: na::SVector<f64, 20> =
         na::SVector::<f64, 20>::from_iterator(log_pi_unormalized.iter().copied());
-    let log_pi = log_pi_unormalized.add_scalar(-FloatTrait::logsumexp(log_pi_unormalized.iter()));
+    let log_pi = log_pi_unormalized.add_scalar(-log_pi_unormalized.iter().ln_sum_exp());
 
     let log_R_mat = {
         let mut mat = na::SMatrix::<f64, 20, 20>::zeros();
@@ -345,7 +356,7 @@ fn rate_matrix(log_R: &[f64], log_pi_unormalized: &[f64]) -> RateMatrixData {
         mat
     };
 
-    let logM = FloatTrait::logsumexp(piRpi.iter());
+    let logM = piRpi.iter().ln_sum_exp();
 
     let logS = {
         let mut mat = na::SMatrix::<f64, 20, 20>::zeros();
@@ -364,7 +375,7 @@ fn rate_matrix(log_R: &[f64], log_pi_unormalized: &[f64]) -> RateMatrixData {
     let sqrt_pi = {
         let mut v = na::SVector::<f64, 20>::zeros();
         for i in 0..20 {
-            v[i] = (log_pi[i] * 0.5).scalar_exp();
+            v[i] = (log_pi[i] * 0.5).exp();
         }
         v
     };
@@ -373,7 +384,7 @@ fn rate_matrix(log_R: &[f64], log_pi_unormalized: &[f64]) -> RateMatrixData {
         let mut mat = na::SMatrix::<f64, 20, 20>::zeros();
         for i in 0..20 {
             for j in 0..20 {
-                mat[(i, j)] = logS[(i, j)].scalar_exp();
+                mat[(i, j)] = logS[(i, j)].exp();
             }
         }
         mat

diff --git a/phylo_grad/Cargo.toml b/phylo_grad/Cargo.toml
@@ -1,20 +1,17 @@
 [package]
 name = "phylo_grad"
-version = "1.1.0"
-edition = "2021"
+version = "2.0.0"
+edition = "2024"
 authors = ["Benjamin Lieser <benjamin.lieser@mpinat.mpg.de>", "Jora Belousov <jorabelousov@gmail.com>"]
 description = "Fast gradient calculation of the Felsenstein algorithm with respect to the rate matrix"
 repository = "https://github.com/soedinglab/phylo_grad"
 license = "MIT OR Apache-2.0"
 keywords = ["math", "felsenstein", "machine-learning", "bioinformatics"]
 categories = ["mathematics", "science::bioinformatics"]
+rust-version = "1.85"
 
 [dependencies]
-logsumexp = "0.1.0"
-nalgebra = "0.33.2"
-num-traits = {version="0.2.19", default-features=false, features = ["libm"]}
+nalgebra = {version ="0.34.1", features = []}
+glam = {version="0.30.10", features = ["fast-math"]}
 rayon = "1.10.0"
-itertools = "0.13.0"
-libm = "0.2.10"
-sleef = "0.3.2"
 faer = {version="0.23.2", default-features=false}
diff --git a/phylo_grad/src/CHANGELOG.md b/phylo_grad/src/CHANGELOG.md
@@ -0,0 +1,15 @@
+# Changelog
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/)
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [2.0.0] — 2026-
+
+### Changes
+
+- The input to phylo_grad are now the probabilties at the leaves and not the log probabilties anymore!
+- The implementation keeps partial likelihoods in linspace internally. It rescales if nessesacry to avoid under and overflows. This leads to a factor 2 speedup, more for the single model case. The output is still a log probability (natural log)
+- We dropped support for f32
+- The parallization of the single model case has been changed to be dramatically more memory efficient.
+- The project does not relay on a nighlty compiler anymore.
Original file line number	Diff line number	Diff line change
		@@ -1 +1 @@
		((((Tip_21:0.6650676023435903,Tip_22:0.024885436124515284):0.2056669529961157,((Tip_37:0.0337728838700196,Tip_38:0.9484958108725318):0.7273031959389747,Tip_36:0.4958302887613361):0.33757637305798605):0.8499954308656537,(Tip_4:0.8141476837951772,Tip_5:0.7402798051460897):0.70741682436221):0.813355,((((((Tip_83:0.17643378492960254,Tip_84:0.7953558100843726):0.14557382727056703,(Tip_43:0.8104945919986606,Tip_44:0.31449340792858616):0.5734673065270548):0.5607844652217767,Tip_18:0.17900274566087643):0.5063065847624981,(((Tip_32:0.6491605888497303,Tip_33:0.3429582140411475):0.14180550845435472,Tip_28:0.08365819037239011):0.5365343800267653,Tip_24:0.8800154043806051):0.998762888840805):0.8304610217387793,(Tip_0:0.7039506434317381,(Tip_25:0.7331478876918618,(Tip_27:0.9441692227658312,((Tip_57:0.38896701363191544,Tip_58:0.13379392008599617):0.8754816391828532,(Tip_60:0.352209387501356,Tip_61:0.3956010586841662):0.4583533711457218):0.5460177635944152):0.6361110453676238):0.6602589686792034):0.027727204798649534):0.6704649990971151,(((Tip_3:0.21104851572654998,(((((((((Tip_73:0.9483105414473861,Tip_74:0.6626788692704468):0.7822936616109846,Tip_59:0.12249378702437955):0.08222970026113245,(Tip_96:0.4290800932938288,Tip_97:0.24924474555329348):0.9179999611656243):0.7548297895532528,(Tip_98:0.9167802235932272,Tip_99:0.7204509004585486):0.8287579648322002):0.05271172049884924,((Tip_81:0.5413926753801398,Tip_82:0.4118750056383439):0.9486624612083683,Tip_39:0.29899889747606284):0.055068995802558016):0.2501971861428551,((Tip_86:0.0380499276137973,Tip_87:0.4256548943232936):0.07123509097099835,((((Tip_55:0.5710466520679875,Tip_56:0.14775864025032526):0.32953437140242836,Tip_47:0.7864549772499915):0.4280637902774943,Tip_45:0.13776127824637646):0.16954936978362714,Tip_23:0.17176851427365566):0.18988101382133601):0.9138259583612118):0.09048036473551212,(((Tip_92:0.404429134512315,Tip_93:0.047246278915062306):0.06224988908172349,((Tip_75:0.6566656102220733,Tip_76:0.6381056061984103):0.29787087139597435,Tip_40:0.5279535103620298):0.8286398644888646):0.19446009582787896,(Tip_34:0.26670755857819356,(Tip_69:0.10832233286814032,Tip_70:0.9365264322108605):0.41936902078708815):0.27879362246002265):0.5567981824116087):0.2694506279899209,Tip_6:0.9312226738371098):0.9215301310000562,(Tip_10:0.4514212322743643,Tip_11:0.18865283598284918):0.8275734698074786):0.23518611049075663):0.992212745439618,((Tip_19:0.8295636279502516,Tip_20:0.34263641161658875):0.9114454035382507,((((Tip_51:0.9220107008370401,Tip_52:0.8176402629230148):0.006951888793413944,(((Tip_94:0.3251911236094175,Tip_95:0.9900649501189612):0.2558363577627372,Tip_85:0.5404121010340436):0.9010280119121413,Tip_29:0.7328316282669919):0.9489623456648526):0.669567713578707,Tip_26:0.8016779879698788):0.6729191308869212,Tip_17:0.29215200135456015):0.31179956452592034):0.2543242891578164):0.5584080530741793,(Tip_66:0.8947516564428961,Tip_67:0.5767430884730821):0.2489036977476457):0.9656347634788244):0.239977,((((Tip_2:0.9464313983922135,(Tip_88:0.41442648404621196,Tip_89:0.5101163097014396):0.5466571358298574):0.6334947392181386,((((((Tip_12:0.11158676296777406,Tip_13:0.7279909147393833):0.2242187756664422,(Tip_41:0.7690871167673865,Tip_42:0.7384233696004908):0.9385579338831107):0.5705174119616021,(Tip_7:0.7360096495183802,Tip_8:0.25431554405864565):0.5988161405333698):0.9687384235590413,((Tip_14:0.49387953934218143,(Tip_62:0.9806096658151946,Tip_63:0.04665946024200317):0.21648740028859653):0.12806441473128458,(((Tip_71:0.4820889062355187,Tip_72:0.669077938124829):0.6907792707402931,Tip_46:0.7915022800264976):0.3224651326307264,Tip_1:0.37807097934208633):0.08262352954341846):0.5415349680132456):0.5368870623401885,(((Tip_9:0.30112516594260985,(Tip_30:0.8010675030911405,Tip_31:0.2197855899426761):0.2979712085110747):0.20561375454872075,(Tip_79:0.322022616870698,Tip_80:0.8405404258659477):0.3736763838229683):0.15489065294790658,((Tip_77:0.1631714880198354,Tip_78:0.14347186085240984):0.6350189977427474,((((Tip_90:0.8211951990555387,Tip_91:0.30263563690395334):0.6502375388726693,Tip_68:0.7812120071752737):0.36830312938864523,Tip_48:0.7430015399338462):0.17280797268648893,Tip_35:0.4635593017282353):0.7149095458288683):0.5852674316858171):0.8996082718667177):0.17732798167953295,((Tip_49:0.33471854305440063,Tip_50:0.11184498641576858):0.8048081839696574,(Tip_53:0.6764255828447392,Tip_54:0.11645530351988755):0.47997626175532904):0.7911201160390272):0.806790252892112):0.43691677945398477,(Tip_64:0.6545202475677102,Tip_65:0.8679242823464907):0.6299303286521293):0.7947232572644647,(Tip_15:0.7247905356618102,Tip_16:0.7397524628980965):0.10514061923847424):1.13067);
		(((Tip_3:0.10805328350042452,Tip_4:0.3315005520372256):0.7775108391547321,(((Tip_10:0.9543578556535617,Tip_11:0.46520848427708783):0.19055443061683935,Tip_5:0.8608787164185184):0.2282830452051495,(Tip_8:0.45877603959910257,Tip_9:0.2188740043797414):0.977427116635245):0.8761171479386695):1.02879,((((Tip_14:0.3495434398130265,Tip_15:0.7072552650535753):0.44843039151645087,(Tip_6:0.8903069753526421,Tip_7:0.11856013796286433):0.6507572690541433):0.7600580845054129,Tip_1:0.2230762517502976):0.04289178403051824,Tip_0:0.7254337418961002):0.143964,((Tip_12:0.535570065080058,Tip_13:0.9288624624322107):0.771915926537751,Tip_2:0.9590719826516791):0.939461);