update

app.py

@@ -16,26 +16,18 @@ logger = logging.getLogger(__name__)
 logger.addHandler(logging.NullHandler())
 
 
-def run_inference(
-    algorithm_version: str,
-    prompt_file: str,
-    prompt_id: int,
-    number_of_samples: int,
-):
+def run_inference(prompt_file: str, prompt_id: int, number_of_samples: int):
 
     # Read file:
     with open(prompt_file.name, "rb") as f:
         prompts = pickle.load(f)
 
-    if all(isinstance(x, str) for x in prompts.keys()):
+    if all(isinstance(x, int) for x in prompts.keys()):
         prompt = prompts[prompt_id]
     else:
         prompt = prompts
 
-    config = GeoDiffGenerator(
-        algorithm_version=algorithm_version,
-        prompt=prompt,
-    )
+    config = GeoDiffGenerator(prompt=prompt)
     model = DiffusersGenerationAlgorithm(config)
     results = list(model.sample(number_of_samples))
     smiles = [Chem.MolToSmiles(m) for m in results]
@@ -45,19 +37,13 @@ def run_inference(
 
 if __name__ == "__main__":
 
-    # Preparation (retrieve all available algorithms)
-    all_algos = ApplicationsRegistry.list_available()
-    algos = [
-        x["algorithm_version"]
-        for x in list(
-            filter(lambda x: "GeoDiff" in x["algorithm_application"], all_algos)
-        )
-    ]
-
     # Load metadata
     metadata_root = pathlib.Path(__file__).parent.joinpath("model_cards")
 
-    examples = [[algos[0], metadata_root.joinpath("mol_dct.pkl"), 2]]
+    examples = [
+        [metadata_root.joinpath("mol_dct.pkl"), 0, 2],
+        [metadata_root.joinpath("mol_dct.pkl"), 1, 2],
+    ]
 
     with open(metadata_root.joinpath("article.md"), "r") as f:
         article = f.read()
@@ -68,9 +54,6 @@ if __name__ == "__main__":
         fn=run_inference,
         title="GeoDiff",
         inputs=[
-            gr.Dropdown(
-                algos, label="GeoDiff version", value="fusing/gfn-molecule-gen-drugs"
-            ),
             gr.File(file_types=[".pkl"], label="GeoDiff prompt"),
             gr.Number(value=0, label="Prompt ID", precision=0),
             gr.Slider(minimum=1, maximum=5, value=2, label="Number of samples", step=1),

model_cards/article.md

@@ -1,32 +1,26 @@
 # Model documentation & parameters
 
-**Diffusion model**: Which model version to use. 
+**GeoDiff prompt**: Here you can upload a `.pkl` file with the necessary variables to initialize a `GeoDiff` generation. Our example file contains five example configurations. NOTE: For details on how to create such files for your custom data, see [original paper](https://openreview.net/forum?id=PzcvxEMzvQC) and this [Colab](https://colab.research.google.com/drive/1pLYYWQhdLuv1q-JtEHGZybxp2RBF8gPs#scrollTo=-3-P4w5sXkRU)
 
-**Prompt**: The text prompt used, only applies to *conditional* diffusion image generators. These are `LDMTextToImageGenerator` and `StableDiffusionGenerator`. The other four models (`DDPMGenerator`, `DDPMGenerator`, `LDMGenerator` and `ScoreSdeGenerator`) are *unconditional*.
+**Prompt ID**: Which of the five example configurations to be used. If you use your own file and have the files in a flat dictionary, leave this blank. If your own file should contain multiple examples, create a top-level dictionary with keys as ascending integers and values as example dictionaries.
 
+**Number of samples**: How many samples should be generated (between 1 and 50).
 
 
-# Model card -- Image diffusion models
 
-**Model Details**: Six diffusion models for image generation:
-- `LDMTextToImageGenerator`
-- `StableDiffusionGenerator`
-- `DDPMGenerator`
-- `DDPMGenerator`
-- `LDMGenerator`
-- `ScoreSdeGenerator`
+# Model card -- GeoDiff
 
-For details, see the [Diffusers docs](https://huggingface.co/docs/diffusers/index)
+**Model Details**: [GeoDiff](https://openreview.net/forum?id=PzcvxEMzvQC): A Geometric Diffusion Model for Molecular Conformation Generation 
 
-**Developers**: Various developers of above models, wrapped by Diffusers developers into [`diffusers`](https://github.com/huggingface/diffusers)
+**Developers**: Minkai Xu and colleagues from MILA and Stanford University.
 
-**Distributors**: Diffusers code integrated into GT4SD.
+**Distributors**: GT4SD Developers.
 
 **Model date**: 2022.
 
-**Model version**: Diffusion models, checkpoints provided and distributed by [`diffusers`](https://github.com/huggingface/diffusers).
+**Model version**: Checkpoints provided by original authors ([see their GitHub repo](https://github.com/MinkaiXu/GeoDiff)).
 
-**Model type**: Various, see [`diffusers`](https://github.com/huggingface/diffusers) docs.
+**Model type**:  A Geometric Diffusion Model for Molecular Conformation Generation 
 
 **Information about training algorithms, parameters, fairness constraints or other applied approaches, and features**: 
 N.A.
@@ -36,11 +30,11 @@ N.A.
 
 **License**: MIT
 
-**Where to send questions or comments about the model**: Open an issue on [`diffusers`](https://github.com/huggingface/diffusers) repo.
+**Where to send questions or comments about the model**: Open an issue on [`GeoDiff`](https://github.com/MinkaiXu/GeoDiff) repo.
 
-**Intended Use. Use cases that were envisioned during development**: Computer vision researchers experimenting with image generative models.
+**Intended Use. Use cases that were envisioned during development**: Chemical research, in particular drug discovery.
 
-**Primary intended uses/users**: Computer vision researchers 
+**Primary intended uses/users**: Researchers and computational chemists using the model for model comparison or research exploration purposes.
 
 **Out-of-scope use cases**: Production-level inference, producing molecules with harmful properties.
 
@@ -56,12 +50,10 @@ Model card prototype inspired by [Mitchell et al. (2019)](https://dl.acm.org/doi
 
 ## Citation
 ```bib
-@misc{von-platen-etal-2022-diffusers,
-  author = {Patrick von Platen and Suraj Patil and Anton Lozhkov and Pedro Cuenca and Nathan Lambert and Kashif Rasul and Mishig Davaadorj and Thomas Wolf},
-  title = {Diffusers: State-of-the-art diffusion models},
-  year = {2022},
-  publisher = {GitHub},
-  journal = {GitHub repository},
-  howpublished = {\url{https://github.com/huggingface/diffusers}}
+@inproceedings{xu2022geodiff,
+  author    = {Minkai Xu and Lantao Yu and Yang Song and Chence Shi and Stefano Ermon and Jian Tang},
+  title     = {GeoDiff: {A} Geometric Diffusion Model for Molecular Conformation Generation},
+  booktitle = {The Tenth International Conference on Learning Representations, {ICLR}},
+  year      = {2022},
 }
 ```

model_cards/description.md

@@ -1,6 +1,6 @@
 <img align="right" src="https://raw.githubusercontent.com/GT4SD/gt4sd-core/main/docs/_static/gt4sd_logo.png" alt="logo" width="120" >
 
-This UI provides access to various diffusion-based image generators implemented in the [`diffusers`](https://github.com/huggingface/diffusers) library, wrapped and re-distributed by GT4SD. 
+[**GeoDiff** (*ICLR* 2022)](https://openreview.net/forum?id=PzcvxEMzvQC) is a diffusion-based molecular generative model implemented in and distributed natively by GT4SD.
 
 For **examples** and **documentation** of the model parameters, please see below.
 Moreover, we provide a **model card** ([Mitchell et al. (2019)](https://dl.acm.org/doi/abs/10.1145/3287560.3287596?casa_token=XD4eHiE2cRUAAAAA:NL11gMa1hGPOUKTAbtXnbVQBDBbjxwcjGECF_i-WC_3g1aBgU1Hbz_f2b4kI_m1in-w__1ztGeHnwHs)) at the bottom of this page.

model_cards/examples.csv

@@ -1,2 +0,0 @@
-LDMGenerator,
-LDMTextToImageGenerator,Generative models on the moon