fix ptable_heatmap_plotly precision keyword renamed to fmt

add figshare URL to pbenner's 16M MACE checkpoint
update pre-commit hooks

Author: janosh

.gitignore

@@ -13,6 +13,7 @@ __pycache__
 data/**/raw
 data/**/tsne
 !data/mp/2023-02-07-mp-elemental-reference-entries.json.gz
+models/**/checkpoints
 
 # slurm + Weights and Biases logs
 wandb/

.pre-commit-config.yaml

@@ -7,7 +7,7 @@ default_install_hook_types: [pre-commit, commit-msg]
 
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.1.2
+    rev: v0.1.5
     hooks:
       - id: ruff
         args: [--fix]
@@ -30,7 +30,7 @@ repos:
       - id: trailing-whitespace
 
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.6.1
+    rev: v1.7.0
     hooks:
       - id: mypy
         additional_dependencies: [types-pyyaml, types-requests]
@@ -56,7 +56,7 @@ repos:
         exclude: ^(site/src/figs/.+\.svelte|data/wbm/20.+\..+|site/src/routes/.+\.(yaml|json)|changelog.md)$
 
   - repo: https://github.com/pre-commit/mirrors-eslint
-    rev: v8.52.0
+    rev: v8.53.0
     hooks:
       - id: eslint
         types: [file]

data/figshare/1.0.0.json

@@ -4,10 +4,14 @@
       "https://figshare.com/ndownloader/files/41233560",
       "2023-06-02-pbenner-best-alignn-model.pth.zip"
     ],
-    "mace_checkpoint": [
+    "mace_checkpoint_1": [
       "https://figshare.com/ndownloader/files/42374049",
       "2023-08-14-mace-yuan-trained-mptrj-04.model"
     ],
+    "mace_checkpoint_2": [
+      "https://figshare.com/ndownloader/files/43117273",
+      "2023-10-29-mace-16M-pbenner-mptrj-no-conditional-loss.model"
+    ],
     "mp_computed_structure_entries": [
       "https://figshare.com/ndownloader/files/40344436",
       "2023-02-07-mp-computed-structure-entries.json.gz"

data/mp/get_mp_energies.py

@@ -111,7 +111,7 @@
 # most points lie on line y=x for x > 0 and y = 0 for x < 0.
 n_above_line = sum(mask_above_line)
 ax.set(
-    title=f"{n_above_line:,} / {len(df):,} = {n_above_line/len(df):.1%} "
+    title=f"{n_above_line:,} / {len(df):,} = {n_above_line / len(df):.1%} "
     "MP materials with\nenergy_above_hull - decomposition_enthalpy.clip(0) > 0.1"
 )
 # ax.figure.savefig(f"{module_dir}/mp-e-above-hull-vs-decomp-enth.webp", dpi=300)

models/cgcnn/train_cgcnn.py

@@ -72,7 +72,7 @@
 for idx in trange(n_perturb, desc="Generating perturbed structures"):
     df_aug[input_col] = [perturb_structure(x) for x in structs]
     df_in = pd.concat(
-        [df_in, df_aug.set_index(f"{x}-aug={idx+1}" for x in df_aug.index)]
+        [df_in, df_aug.set_index(f"{x}-aug={idx + 1}" for x in df_aug.index)]
     )
 
 del df_aug

models/mace/test_mace.py

@@ -39,21 +39,22 @@
 relax_cell = True
 model_name = [
     # MACE trained on M3GNet training set by original MACE authors
-    "2023-07-14-mace-ilyes-MPF-2021-2-8-big-128-6",
+    # 9M = model size (number of trainable params), 9,255,168 to be exact
+    "2023-07-14-mace-9M-ilyas-universal-2-big-128-6",
     # MACE trained by Yuan Chiang on CHGNet training set
-    "2023-08-14-mace-yuan-mptrj-04",
-    "2023-09-03-mace-yuan-mptrj-slower-14-lr-13_run-3",
-    "2023-10-29-mace-pbenner-mptrj-no-conditional-loss",
-][-1]
+    "2023-08-14-mace-2M-yuan-mptrj-04",
+    "2023-09-03-mace-2M-yuan-mptrj-slower-14-lr-13_run-3",
+    "2023-10-29-mace-16M-pbenner-mptrj-no-conditional-loss",
+][0]
 
 slurm_vars = slurm_submit(
     job_name=job_name,
     out_dir=out_dir,
     account="matgen",
-    time="4:55:0",
+    time="9:55:0",
     array=f"1-{slurm_array_task_count}",
-    # slurm_flags="--qos shared --constraint gpu --gpus 1",
-    slurm_flags="--qos shared --constraint cpu --mem 16G",
+    slurm_flags="--qos shared --constraint gpu --gpus 1",
+    # slurm_flags="--qos shared --constraint cpu --mem 16G",
 )
 
 

models/voronoi/train_test_voronoi_rf.py

@@ -90,7 +90,7 @@
 feature_names = featurizer.feature_labels()
 n_nans = df_train[feature_names].isna().any(axis=1).sum()
 
-print(f"train set NaNs: {n_nans:,} / {len(df_train):,} = {n_nans/len(df_train):.3%}")
+print(f"train set NaNs: {n_nans:,} / {len(df_train):,} = {n_nans / len(df_train):.3%}")
 
 df_train = df_train.dropna(subset=feature_names)
 
@@ -110,7 +110,7 @@
 
 # %%
 n_nans = df_test[feature_names].isna().any(axis=1).sum()
-print(f"test set NaNs: {n_nans:,} / {len(df_train):,} = {n_nans/len(df_train):.1%}")
+print(f"test set NaNs: {n_nans:,} / {len(df_train):,} = {n_nans / len(df_train):.1%}")
 
 df_test = df_test.dropna(subset=feature_names)
 

scripts/analyze_model_failure_cases.py

@@ -197,7 +197,7 @@
         if normalized
         else elem_counts[col],
         color_bar=dict(title=col),
-        precision=".3f",
+        fmt=".3f",
         cscale_range=[0, 0.1],
     )
     fig.show()

scripts/model_figs/per_element_errors.py

@@ -116,7 +116,7 @@
 
 # %% plot per-element std dev of DFT hull dist
 fig = ptable_heatmap_plotly(
-    df_elem_err[test_set_std_col], precision=".2f", colorscale="Inferno"
+    df_elem_err[test_set_std_col], fmt=".2f", colorscale="Inferno"
 )
 fig.show()
 
@@ -136,7 +136,7 @@
         per_elem_err /= df_elem_err[test_set_std_col]
         per_elem_err.name = f"{model} (normalized by test set std)"
     fig = ptable_heatmap_plotly(
-        per_elem_err, precision=".2f", colorscale="Inferno", cscale_range=cs_range
+        per_elem_err, fmt=".2f", colorscale="Inferno", cscale_range=cs_range
     )
     fig.show()
 

site/src/routes/contribute/+page.md

@@ -10,7 +10,8 @@
 
   const descriptions = {
     alignn_checkpoint: "ALIGNN model trained on <code>mp_computed_structure_entries</code>",
-    mace_checkpoint: "MACE model trained on the MPtrj dataset (https://figshare.com/articles/dataset/23713842)",
+    mace_checkpoint_1: "2M params MACE model trained by Yuan Chiang on MPtrj (https://figshare.com/articles/dataset/23713842)",
+    mace_checkpoint_2: "16M params MACE model trained by Philipp Benner on MPtrj (https://figshare.com/articles/dataset/23713842)",
     mp_computed_structure_entries:
       `JSON-Serialized MP ${cse_link} objects containing relaxed structures and DFT final energies`,
     mp_elemental_ref_entries: `Minimum energy ComputedEntry for each element in MP`,