Merge branch 'DOR-926-polish-model-names' into 'master'

Dorado Polish: fix the model naming convention See merge request machine-learning/dorado!1314
nanoporetech · Dec 16, 2024 · 8e0e478 · 8e0e478
2 parents 492ef84 + f507d2f
commit 8e0e478
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 18 deletions.
diff --git a/dorado/cli/polish.cpp b/dorado/cli/polish.cpp
@@ -588,9 +588,6 @@ const polisher::ModelConfig resolve_model(const polisher::BamInfo& bam_info,
         // Example: dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl_mv
         std::string model_name = basecaller_model + polish_model_suffix;
 
-        // Example: dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl_mv
-        std::replace(std::begin(model_name), std::end(model_name), '@', '_');
-
         spdlog::info("Downloading model: '{}'", model_name);
         model_dir = download_model(model_name);
 
@@ -613,9 +610,6 @@ const polisher::ModelConfig resolve_model(const polisher::BamInfo& bam_info,
         // Example: dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl_mv
         std::string model_name = basecaller_model + polish_model_suffix;
 
-        // Example: dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl_mv
-        std::replace(std::begin(model_name), std::end(model_name), '@', '_');
-
         spdlog::info("Downloading model: '{}'", model_name);
         model_dir = download_model(model_name);
 

diff --git a/dorado/models/models.cpp b/dorado/models/models.cpp
@@ -1121,29 +1121,29 @@ namespace polisher {
 
 const std::vector<ModelInfo> models = {
         ModelInfo{
-                "dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl",
-                "fdb6d18fe7e93f141c8e15f1b014e338e7cdd2b3f390abd0ff85cede33b50cc7",
+                "dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl",
+                "d343b4394b904d219257ad188c82ece63b935f15d78f09f551e591b2275da4b9",
                 CC::UNKNOWN,
                 ModelVariantPair{},
                 ModsVariantPair{},
         },
         ModelInfo{
-                "dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl_mv",
-                "719389cf723a4ddbd020690c20708885ed6b5000c51a2144a5af5b8064a055d2",
+                "dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl_mv",
+                "928d9bcf3d68162eff479ada5839c5df3faa0ad393658729511aedffe65f089c",
                 CC::UNKNOWN,
                 ModelVariantPair{},
                 ModsVariantPair{},
         },
         ModelInfo{
-                "dna_r10.4.1_e8.2_400bps_sup_v5.0.0_polish_rl",
-                "3c0c7fbf2f204390ca6dfafaefc25a3f999870368e4a7b5d238eb600aa7e781b",
+                "dna_r10.4.1_e8.2_400bps_sup@v5.0.0_polish_rl",
+                "6d8c5a8ce45311c25f824453d0af997fbe2f63a5f734fdb4d884d285ddafec33",
                 CC::UNKNOWN,
                 ModelVariantPair{},
                 ModsVariantPair{},
         },
         ModelInfo{
-                "dna_r10.4.1_e8.2_400bps_sup_v5.0.0_polish_rl_mv",
-                "eb9b4691bda3133e9b991aa1a2666cc632bbd5c2f803dff704ef1265f4315a90",
+                "dna_r10.4.1_e8.2_400bps_sup@v5.0.0_polish_rl_mv",
+                "0e0cb175aa41636de835d2abb5330b91fed14a00f811804edf983bc086cf477a",
                 CC::UNKNOWN,
                 ModelVariantPair{},
                 ModsVariantPair{},

diff --git a/tests/cram/polish/cram-polish-01-model-download.t b/tests/cram/polish/cram-polish-01-model-download.t
@@ -27,8 +27,8 @@ Negative test: BAM has a model which is not available for download in auto mode.
   > echo "Exit code: $?"
   > grep "\[error\]" out/out.fasta.stderr | sed -E 's/.*\[error\] //g'
   Exit code: 1
-  Selected model doesn't exist: dna_r10.4.1_e8.2_400bps_hac_v1.0.0_polish_rl_mv
-  Could not download model: dna_r10.4.1_e8.2_400bps_hac_v1.0.0_polish_rl_mv
+  Selected model doesn't exist: dna_r10.4.1_e8.2_400bps_hac@v1.0.0_polish_rl_mv
+  Could not download model: dna_r10.4.1_e8.2_400bps_hac@v1.0.0_polish_rl_mv
 
 Negative test: BAM has no models listed (no RG tags).
   $ rm -rf out; mkdir -p out
@@ -101,7 +101,7 @@ Load a model from path.
 Download an explicit polishing model by name with `--skip-model-compatibility-check`.
   $ rm -rf out; mkdir -p out
   > in_dir=${TEST_DATA_DIR}/polish/test-01-supertiny
-  > model="dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl_mv"
+  > model="dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl_mv"
   > ${DORADO_BIN} polish --skip-model-compatibility-check --model "${model}" --device cpu ${in_dir}/calls_to_draft.bam ${in_dir}/draft.fasta.gz -t 4 --infer-threads 1 -vv > out/out.fasta 2> out/out.fasta.stderr
   > echo "Exit code: $?"
   > ${DORADO_BIN} aligner ${in_dir}/ref.fasta.gz out/out.fasta 1> out/out.sam 2> out/out.sam.stderr

diff --git a/tests/test_dorado_polish.sh b/tests/test_dorado_polish.sh
@@ -49,7 +49,7 @@ CRAM=$(pwd)/cram-0.6/cram.py
 popd
 
 # Download the model once.
-MODEL_NAME="dna_r10.4.1_e8.2_400bps_hac_v5.0.0_polish_rl_mv"
+MODEL_NAME="dna_r10.4.1_e8.2_400bps_hac@v5.0.0_polish_rl_mv"
 MODEL_DIR=${output_dir}/${MODEL_NAME}
 if [[ ! -d "${MODEL_DIR}" ]]; then
     ${DORADO_BIN} download --model "${MODEL_NAME}" --models-directory ${output_dir}