Enable GTR benchmarking in treeflow

- add parse-trace.py
- use -with-trace in nextflow

Dockerfile

@@ -66,8 +66,7 @@ RUN cd /phylojax && /opt/conda/envs/bito/bin/pip install jax==0.2.24 jaxlib . &&
 RUN ln -s /phylojax/benchmarks/benchmark.py /usr/local/bin/phylojax-benchmark \
     && chmod +x /usr/local/bin/phylojax-benchmark
 
-RUN git clone --depth 1 https://github.com/4ment/phylostan /phylostan
-RUN cd /phylostan && pip install . && phylostan --help
+RUN pip install phylostan==1.0.5 && phylostan --help
 
 RUN git clone --depth 1 https://github.com/christiaanjs/treeflow.git /treeflow
 RUN cd /treeflow && /opt/conda/envs/bito/bin/pip install . 

configs/uts.config

@@ -24,6 +24,7 @@ process{
         time = '5m'
     }
     withName: RUN_TREEFLOW{
+        clusterOptions = '-l select=1:ncpus=3'
         time = '20h'
     }
 }

flu_H3N2/physher-JC69.template

@@ -128,6 +128,7 @@ RATE_PRIOR_TEMPLATE
 		"tol": TOL_TEMPLATE,
 		"eta": LR_TEMPLATE,
 		"model": "&varnormal",
+		"update": "adam",
 		"checkpoint": "CHECKPOINT_TEMPLATE",
 		"parameters": ["%mu", "%sigma", "&mu.rate", "&sigma.rate","&mu.theta", "&sigma.theta"],
 		"max": TEMPLATE_ITER

modules/micro.nf

@@ -94,6 +94,8 @@ process RUN_TREEFLOW_BENCHMARK {
                       -t $lsd_newick \
                       -r ${params.replicates} \
                       -s 0.001 \
+                      --gtr \
+                      -p false \
                       -o out.csv
   awk 'NR==1{print "program,size,rep,precision,"\$0}; \
        NR>1{print "treeflow,$size,$rep,64,"\$0}' out.csv \

run_rhino.sh

@@ -14,4 +14,5 @@ export PATH=$SINGULARITYROOT/bin/:$PATH
     -profile rhino \
     -with-report ./r_output/nextflow_report.html \
     -work-dir ./r_output/work/ \
+    -with-trace \
     -resume

scripts/parse-trace.py

@@ -0,0 +1,39 @@
+import sys
+from os import walk
+
+if len(sys.argv) == 1:
+    print(f"Usage: python {sys.argv[0]} work/ trace.txt")
+    exit(2)
+
+dic = {}
+for (dirpath, dirnames, filenames) in walk(sys.argv[1]):
+    if len(filenames) > 0:
+        dic[dirpath.lstrip("work/")[:9]] = filenames
+
+with open(sys.argv[2], "r") as fp:
+    for line in fp:
+        a = line.split("\t")
+        if "task_id" == a[0]:
+            print("program\tsize\treplicate\t" + line, end="")
+        elif "macro_flu:RUN_" in a[3]:
+            for f in dic[a[1]]:
+                if ".txt" in f:
+                    b = f.split(".")[:-1]
+                    mem = a[10].split(" ")
+                    if mem[1] == "GB":
+                        a[10] = str(float(mem[0]) * 1024)
+                    else:
+                        a[10] = mem[0]
+                    mem = a[11].split(" ")
+                    if mem[1] == "GB":
+                        a[11] = str(float(mem[0]) * 1024)
+                    else:
+                        a[11] = mem[0]
+
+                    if len(b) == 4:
+                        b[0] = "bitorch" if b[1] == "true" else "torchtree"
+                        del b[1]
+                        print("\t".join(b) + "\t" + "\t".join(a), end="")
+                    else:
+                        print("\t".join(b) + "\t" + "\t".join(a), end="")
+                    break