murrellb commited on
Commit
7b37970
·
0 Parent(s):

Molecular LPGM reference models

Browse files
Files changed (7) hide show
  1. .gitattributes +35 -0
  2. README.md +43 -0
  3. checkpoints.json +43 -0
  4. geom/bm.pt +3 -0
  5. geom/fragment.pt +3 -0
  6. qm9/bm.pt +3 -0
  7. qm9/fragment.pt +3 -0
.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: pytorch
3
+ tags:
4
+ - molecular-generation
5
+ - flow-matching
6
+ - geom
7
+ - qm9
8
+ ---
9
+ # Molecular LPGM
10
+
11
+ [Reference implementation and documentation](https://github.com/MurrellGroup/molLPGM)
12
+
13
+ Final **Fragment Noise** and **Brownian Atoms** models for GEOM-Drugs and QM9.
14
+ Both methods condition on atom count and the BRICS fragment partition, with
15
+ learned same/different-fragment attention biases in every transformer layer.
16
+ Neither receives clean coordinates, atom labels, bonds or geometric templates
17
+ at generation. Fragment Noise uses a single fixed global angular variance of
18
+ **6** in both datasets.
19
+
20
+ | Dataset | Width | Layers | Fragment Noise | Brownian Atoms |
21
+ | --- | ---: | ---: | --- | --- |
22
+ | GEOM-Drugs | 768 | 16 | [fragment.pt](geom/fragment.pt) | [bm.pt](geom/bm.pt) |
23
+ | QM9 | 512 | 12 | [fragment.pt](qm9/fragment.pt) | [bm.pt](qm9/bm.pt) |
24
+
25
+ All four checkpoints contain final **800,000-update raw and EMA weights**
26
+ (decay **0.9999**), architecture, process parameters, and final loss weights.
27
+ The final 100,000 updates are learning-rate warmdown. Optimizer state is omitted.
28
+ Use `--ema` in the reference sampling commands, or `load_checkpoint(...,
29
+ use_ema=True)` in Python. SHA-256 hashes and parameter counts are in
30
+ [checkpoints.json](checkpoints.json).
31
+
32
+ The source is a centered Gaussian with standard deviation 3 Å. Brownian Atoms
33
+ uses the scalar clock `u(t) = 1 - (1 - t**2.3)**3.1` and variance coefficient
34
+ `beta = 14.8 Ų`; these parameters approximately match the conditional
35
+ endpoint spread. It has one Cartesian coordinate head and the same atom-label
36
+ process as Fragment Noise.
37
+
38
+ [10K-sample results and archives](https://github.com/MurrellGroup/molLPGM/tree/main/results)
39
+ use final EMA weights and 2,000 square-root-spaced steps. The same ADiT validity,
40
+ valid-and-connected, and uniqueness protocol is used for both datasets.
41
+ GEOM evaluation uses test cases; QM9 pools validation and test cases.
42
+ See the [comparison protocol](https://github.com/MurrellGroup/molLPGM/blob/main/docs/comparison.md)
43
+ for conditioning, objective and training details.
checkpoints.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "repository": "MurrellLab/LPGM",
3
+ "geom": {
4
+ "fragment": {
5
+ "file": "geom/fragment.pt",
6
+ "sha256": "de537eaba256c5c3d8cb14a879546666d614af405aaf73ccafec2a4f579e5859",
7
+ "bytes": 1731183363,
8
+ "parameters": 216378263,
9
+ "step": 800000,
10
+ "ema_decay": 0.9999,
11
+ "url": "https://huggingface.co/MurrellLab/LPGM/resolve/main/geom/fragment.pt"
12
+ },
13
+ "bm": {
14
+ "file": "geom/bm.pt",
15
+ "sha256": "b4a381cb7ecb311f4c351f3c62c7a54aa12b08a6e8849da6b200b560215c9b21",
16
+ "bytes": 1731132939,
17
+ "parameters": 216375953,
18
+ "step": 800000,
19
+ "ema_decay": 0.9999,
20
+ "url": "https://huggingface.co/MurrellLab/LPGM/resolve/main/geom/bm.pt"
21
+ }
22
+ },
23
+ "qm9": {
24
+ "fragment": {
25
+ "file": "qm9/fragment.pt",
26
+ "sha256": "87c22cc549ae3a5a7f1dedddcabae1ddd69c12c126e7fc6fc82f504fd9a747e2",
27
+ "bytes": 588854123,
28
+ "parameters": 73591607,
29
+ "step": 800000,
30
+ "ema_decay": 0.9999,
31
+ "url": "https://huggingface.co/MurrellLab/LPGM/resolve/main/qm9/fragment.pt"
32
+ },
33
+ "bm": {
34
+ "file": "qm9/bm.pt",
35
+ "sha256": "fb431bca7398c6c0dd6fa27ffa8e03797533c3feff696743416b15fc7f9e2582",
36
+ "bytes": 588816611,
37
+ "parameters": 73590065,
38
+ "step": 800000,
39
+ "ema_decay": 0.9999,
40
+ "url": "https://huggingface.co/MurrellLab/LPGM/resolve/main/qm9/bm.pt"
41
+ }
42
+ }
43
+ }
geom/bm.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4a381cb7ecb311f4c351f3c62c7a54aa12b08a6e8849da6b200b560215c9b21
3
+ size 1731132939
geom/fragment.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de537eaba256c5c3d8cb14a879546666d614af405aaf73ccafec2a4f579e5859
3
+ size 1731183363
qm9/bm.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb431bca7398c6c0dd6fa27ffa8e03797533c3feff696743416b15fc7f9e2582
3
+ size 588816611
qm9/fragment.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87c22cc549ae3a5a7f1dedddcabae1ddd69c12c126e7fc6fc82f504fd9a747e2
3
+ size 588854123