update

app.py

@@ -1,142 +1,122 @@
 import gradio as gr
-import numpy as np
-import random
-#import spaces #[uncomment to use ZeroGPU]
-from diffusers import DiffusionPipeline
-import torch
+from gradio_molecule2d import molecule2d
+from synformer.chem.mol import Molecule
+from synformer.sampler.analog.parallel import run_sampling_one_cpu
+from huggingface_hub import hf_hub_download
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
-model_repo_id = "stabilityai/sdxl-turbo" #Replace to the model you would like to use
+REPO_ID = "whgao/synformer"
+CKPT_FILENAME = "sf_ed_default.ckpt"
+MAT_FILENAME = "matrix.pkl"
+FPI_FILENAME = "fpindex.pkl"
+ckpt_path = hf_hub_download(REPO_ID, CKPT_FILENAME)
+mat_path = hf_hub_download(REPO_ID, MAT_FILENAME)
+fpi_path = hf_hub_download(REPO_ID, FPI_FILENAME)
 
-if torch.cuda.is_available():
-    torch_dtype = torch.float16
-else:
-    torch_dtype = torch.float32
+last_result = {}
 
-pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype)
-pipe = pipe.to(device)
+# Function to clear all inputs
+def clear_inputs():
+    # Return default or empty values to reset each input component
+    return None, 24, 64, 0
 
-MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
+def sample(smi, search_width, exhaustiveness):
+    result_df = run_sampling_one_cpu(
+        input=Molecule(smi),
+        model_path=ckpt_path,
+        mat_path=mat_path,
+        fpi_path=fpi_path,
+        search_width=search_width,
+        exhaustiveness=exhaustiveness,
+        time_limit=180,
+        max_results=100,
+        max_evolve_steps=24,
+        sort_by_scores=True,
+    )
+    result_df = result_df[:30]
+    last_result["results_df"] = result_df
+    smiles = result_df.iloc[0]["smiles"]
+    similarity = result_df.iloc[0]["score"]
+    synthesis = result_df.iloc[0]["synthesis"]
+    return smiles, similarity, synthesis, gr.update(maximum=len(result_df)-1)
 
-#@spaces.GPU #[uncomment to use ZeroGPU]
-def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
 
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-        
-    generator = torch.Generator().manual_seed(seed)
-    
-    image = pipe(
-        prompt = prompt, 
-        negative_prompt = negative_prompt,
-        guidance_scale = guidance_scale, 
-        num_inference_steps = num_inference_steps, 
-        width = width, 
-        height = height,
-        generator = generator
-    ).images[0] 
-    
-    return image, seed
+def select_from_output(index):
+    df_results = last_result["results_df"]
+    return df_results.iloc[index]["smiles"], df_results.iloc[index]["score"], df_results.iloc[index]["synthesis"]
 
 examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
+    "Nc1cccc(S(=O)(=O)N2CCCN(S(=O)(=O)c3ccc4c(c3)OCCO4)CC2)c1",
+    "CN1C[C@H](Nc2cnn(C)c(=O)c2)C[C@H](c2ccccc2)C1",
+    "COc1ccc(-c2ccnc(Nc3ccccc3)n2)cc1",
+    "CC[C@@H]1OC[C@@]23Cc4cc(F)c(N)cc4-c4ccc5c(c42)C(=CC(F)(F)O5)[C@@H]1C3=O",
+    "O=C(OCC(=O)N1[C@H](C(=O)O)C[C@@H]2CCCC[C@@H]21)[C@H](Cc1cbccc1)NC(I)c1bcccc1",
 ]
 
-css="""
-#col-container {
-    margin: 0 auto;
-    max-width: 640px;
-}
-"""
 
-with gr.Blocks(css=css) as demo:
-    
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown(f"""
-        # Text-to-Image Gradio Template
+with gr.Blocks() as demo:
+    gr.Markdown(f"""
+        # Demo of [SynFormer](https://github.com/wenhao-gao/synformer/tree/main)
+        This page demonstrates the SynFormer-ED model, which takes a molecule as input—regardless of its synthetic accessibility—and outputs 
+                identical or approximate molecules along with their associated synthetic paths. The demo runs on CPUs and typically takes about 
+                one minute per run but can be accelerated by reducing the search width and exhaustiveness. The model may take longer if the server 
+                is busy. Since the sampling is stochastic, you may run the demo multiple times to explore different results, with a maximum of 
+                30 molecules displayed at once.
+        To learn more about SynFormer’s architecture and applications, check out [our paper](https://github.com/wenhao-gao/synformer/tree/main).
+                
+        Authors: [Wenhao Gao](mailto:[email protected]), Shitong Luo, Connor W. Coley
         """)
-        
-        with gr.Row():
-            
-            prompt = gr.Text(
-                label="Prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter your prompt",
-                container=False,
-            )
-            
-            run_button = gr.Button("Run", scale=0)
-        
-        result = gr.Image(label="Result", show_label=False)
+    with gr.Row():
+        with gr.Column(scale=0.5):
+            input_molecule = molecule2d(label="SMILES Input")
+            slider_1 = gr.Slider(minimum=1, maximum=100, step=1, label="Search Width", value=24)
+            slider_2 = gr.Slider(minimum=1, maximum=100, step=1, label="Exhaustiveness", value=64)
 
-        with gr.Accordion("Advanced Settings", open=False):
-            
-            negative_prompt = gr.Text(
-                label="Negative prompt",
-                max_lines=1,
-                placeholder="Enter a negative prompt",
-                visible=False,
-            )
-            
-            seed = gr.Slider(
-                label="Seed",
-                minimum=0,
-                maximum=MAX_SEED,
-                step=1,
-                value=0,
-            )
-            
-            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-            
-            with gr.Row():
-                
-                width = gr.Slider(
-                    label="Width",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024, #Replace with defaults that work for your model
-                )
-                
-                height = gr.Slider(
-                    label="Height",
-                    minimum=256,
-                    maximum=MAX_IMAGE_SIZE,
-                    step=32,
-                    value=1024, #Replace with defaults that work for your model
-                )
-            
             with gr.Row():
-                
-                guidance_scale = gr.Slider(
-                    label="Guidance scale",
-                    minimum=0.0,
-                    maximum=10.0,
-                    step=0.1,
-                    value=0.0, #Replace with defaults that work for your model
-                )
-                
-                num_inference_steps = gr.Slider(
-                    label="Number of inference steps",
-                    minimum=1,
-                    maximum=50,
-                    step=1,
-                    value=2, #Replace with defaults that work for your model
-                )
-        
-        gr.Examples(
-            examples = examples,
-            inputs = [prompt]
-        )
-    gr.on(
-        triggers=[run_button.click, prompt.submit],
-        fn = infer,
-        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
-        outputs = [result, seed]
+                with gr.Column(scale=0.5):
+                    run_btn = gr.Button("Run on sample")
+                with gr.Column(scale=0.5):
+                    clear_btn = gr.Button("Clear")
+
+        with gr.Column(scale=0.5):
+            index_slider = gr.Slider(minimum=0, maximum=10, step=1, label="Select Output Index", value=0, interactive=True) 
+            output_similarity = gr.Text(label="Tanimoto Similarity")
+            output_molecule = molecule2d(label="Output")
+            output_synpath = gr.Textbox(label="Synthetic Path")
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.Markdown("### Examples")
+            gr.Examples(
+                examples = examples,
+                inputs = [input_molecule]
+            )
+    
+    run_btn.click(
+        fn=sample, 
+        inputs=[
+            input_molecule, 
+            slider_1, 
+            slider_2
+        ], 
+        outputs=[
+            output_molecule, 
+            output_similarity,
+            output_synpath,
+            index_slider
+        ],
+        api_name="Run"
+    )
+
+    index_slider.change(
+        fn=select_from_output,
+        inputs=[index_slider],
+        outputs=[output_molecule, output_similarity, output_synpath],
+    )
+
+    clear_btn.click(
+        fn=clear_inputs,
+        inputs=[],
+        outputs=[input_molecule, slider_1, slider_2, index_slider]
     )
 
-demo.queue().launch()
+demo.launch()

requirements.txt

@@ -1,6 +1,4 @@
-accelerate
-diffusers
-invisible_watermark
+gradio_molecule2d
 torch
-transformers
-xformers
+synformer
+huggingface-hub