Token Classification
Transformers
PEFT
PyTorch
protein-language-model
protein-annotation
bioinformatics
esm2
Instructions to use Wangwpi/PALM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Wangwpi/PALM with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("token-classification", model="Wangwpi/PALM")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Wangwpi/PALM", device_map="auto") - PEFT
How to use Wangwpi/PALM with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -24,9 +24,23 @@ pathogenicity/random_split_fold_2.pt
|
|
| 24 |
pathogenicity/random_split_fold_3.pt
|
| 25 |
pathogenicity/random_split_fold_4.pt
|
| 26 |
pathogenicity/random_split_fold_5.pt
|
|
|
|
| 27 |
checkpoint_manifest.json
|
| 28 |
```
|
| 29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
The annotation checkpoint predicts four residue-level outputs:
|
| 31 |
|
| 32 |
- domain occurrence: none, one, or multiple;
|
|
@@ -53,8 +67,9 @@ predictor = PALMPredictor(
|
|
| 53 |
annotation = predictor.annotate("MSEQUENCE...")
|
| 54 |
```
|
| 55 |
|
| 56 |
-
The checkpoint is downloaded from the
|
| 57 |
-
|
|
|
|
| 58 |
|
| 59 |
## Pathogenicity heads
|
| 60 |
|
|
@@ -76,7 +91,6 @@ separately without loading these heads.
|
|
| 76 |
- Cohort filters and transcript aggregation must be reported with benchmark
|
| 77 |
results.
|
| 78 |
|
| 79 |
-
## Citation
|
| 80 |
|
| 81 |
-
|
| 82 |
-
license before publishing this model repository.
|
|
|
|
| 24 |
pathogenicity/random_split_fold_3.pt
|
| 25 |
pathogenicity/random_split_fold_4.pt
|
| 26 |
pathogenicity/random_split_fold_5.pt
|
| 27 |
+
interpro_classes.tsv
|
| 28 |
checkpoint_manifest.json
|
| 29 |
```
|
| 30 |
|
| 31 |
+
`interpro_classes.tsv` maps every output class to its InterPro entry:
|
| 32 |
+
|
| 33 |
+
```text
|
| 34 |
+
head class_index interpro_id interpro_type name
|
| 35 |
+
domain 0 IPR000001 Domain Kringle
|
| 36 |
+
site 0 IPR000035 Conserved_site Alkylbase DNA glycosidase, conserved site
|
| 37 |
+
```
|
| 38 |
+
|
| 39 |
+
`class_index` is 0-based and per head: domain indices run 0-17950, and site
|
| 40 |
+
indices restart at 0 and run to 1382. The checkpoint itself carries the same
|
| 41 |
+
accession-to-index maps as `ipr2domain` and `ipr2site`; this file adds the
|
| 42 |
+
human-readable names.
|
| 43 |
+
|
| 44 |
The annotation checkpoint predicts four residue-level outputs:
|
| 45 |
|
| 46 |
- domain occurrence: none, one, or multiple;
|
|
|
|
| 67 |
annotation = predictor.annotate("MSEQUENCE...")
|
| 68 |
```
|
| 69 |
|
| 70 |
+
The checkpoint is downloaded from the
|
| 71 |
+
[`Wangwpi/PALM`](https://huggingface.co/Wangwpi/PALM) model repository and
|
| 72 |
+
cached by `huggingface_hub`.
|
| 73 |
|
| 74 |
## Pathogenicity heads
|
| 75 |
|
|
|
|
| 91 |
- Cohort filters and transcript aggregation must be reported with benchmark
|
| 92 |
results.
|
| 93 |
|
| 94 |
+
## Citation
|
| 95 |
|
| 96 |
+
Citation information will be added when the PALM manuscript is available.
|
|
|