docs: revision (#43)

Author: leonwehrhan

docs/source/benchmarks/general/scaling.rst

@@ -7,7 +7,7 @@ Purpose
 -------
 
 This benchmark evaluates how the computational cost of machine-learned interatomic potentials (**MLIP**) scales with system size.
-By running single, long **MD** episodes on a series of molecular systems of increasing size, we systematically assess the
+By running single **MD** episodes on a series of molecular systems of increasing size, we systematically assess the
 relationship between molecular complexity and inference performance. The results provide insight into the efficiency and
 scalability of the **MLIP** implementation, helping to identify potential bottlenecks and guide optimization for large-scale
 simulations.
@@ -16,7 +16,7 @@ Description
 -----------
 
 For each system in the dataset, the benchmark performs a **MD** simulation using the  **MLIP** model in the **NVT** ensemble at **300 K**
-for **1,000,000 steps** (1ns), leveraging the `jax-md <https://github.com/google/jax-md>`_, as integrated via the
+for **1000 steps** (1 ps), leveraging the `jax-md <https://github.com/google/jax-md>`_, as integrated via the
 `mlip <https://github.com/instadeepai/mlip>`_ library. During each simulation, a timer tracks the duration of each episode, and the average episode time (excluding the first episode)
 is recorded. After all simulations are complete, the benchmark reports the **average inference time per episode as a function of
 system size**, providing a direct measure of how the **MLIP** implementation's computational cost grows with increasing molecular
@@ -46,4 +46,4 @@ They have the following ids:
 Interpretation
 --------------
 
-The inference scaling score is a measure of the scaling of the **MLIP** model.
+This benchmark does not produce a score but can be used to estimate how a model's simulation speed scales with system size.

docs/source/benchmarks/general/stability.rst

@@ -13,7 +13,7 @@ Description
 -----------
 
 For each system in the dataset, the benchmark performs a **MD** simulation using the  **MLIP** model in the
-**NVT** ensemble at **300 K** for **1,000,000 steps** (1ns), leveraging the
+**NVT** ensemble at **300 K** for **100,000 steps** (100 ps), leveraging the
 `jax-md <https://github.com/google/jax-md>`_, as integrated via the `mlip <https://github.com/instadeepai/mlip>`_
 library. The test monitors the system for signs of instability by detecting abrupt temperature spikes
 (**“explosions”**) and hydrogen atom drift. These indicators help determine whether the **MLIP** maintains
@@ -39,15 +39,14 @@ Dataset
 The structures that are tested for stability are a series of protein structures, RNA fragments, peptides and inhibitors taken from the PDB.
 They have the following ids:
 
-* 1JRS
-* 1UAO
-* 1P79
-* 5KGZ
-* 1AB7
-* 1BIP
-* 1A5E
-* 1A7M
-* 2BQV
+* 1JRS (Leupeptin)
+* 1UAO (Chignolin)
+* 1P79 (RNA Fragment)
+* 5KGZ (Protein structure with 634 atoms)
+* 1AB7 (Protein structure with 1,432 atoms)
+* 1BIP (Protein structure with 1,818 atoms)
+* 1A5E (Protein structure with 2,301 atoms)
+* 1A7M (Protein structure with 2,803 atoms)
 
 Interpretation
 --------------

docs/source/benchmarks/small_molecules/conformer_selection.rst

@@ -21,9 +21,7 @@ For each system, the benchmark leverages the `mlip <https://github.com/instadeep
 comparing the predicted energies and forces against quantum mechanical **QM** reference data. Performance is quantified using
 the following metrics:
 
-- **MAE (Mean Absolute Error)** and **RMSE (Root Mean Square Error)** for:
-  - Total energies (in kcal/mol)
-  - Atomic forces (in kcal/mol/Å)
+- **MAE (Mean Absolute Error)** and **RMSE (Root Mean Square Error)** for total energies (in kcal/mol)
 - **Spearman rank correlation coefficient** for conformer energy ordering
 
 

docs/source/benchmarks/small_molecules/dihedral_scan.rst

@@ -31,7 +31,6 @@ Performance is quantified using the following metrics:
           :align: center
           :figclass: align-center
 
-          Structure 1
      - .. figure:: img/dihedral_scan.png
           :width: 100%
           :align: center

docs/source/benchmarks/small_molecules/tautomers.rst

@@ -27,8 +27,8 @@ For each molecule, the benchmark compares **MLIP**-predicted energies against
 quantum mechanical **QM** reference data. Performance
 is quantified using the following metrics:
 
-- :abbr:`MAE (Mean Absolute Error)`
-- :abbr:`RMSE (Root Mean Square Error)`
+- **MAE (Mean Absolute Error)**
+- **RMSE (Root Mean Square Error)**
 
 
 Dataset