File size: 4,900 Bytes
c800551
 
 
 
cb662f9
c800551
eb03c6d
c800551
a049b66
 
 
c800551
cb662f9
a049b66
eb03c6d
 
a049b66
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eb03c6d
cb662f9
eb03c6d
 
a049b66
eb03c6d
cb662f9
eb03c6d
 
a049b66
f966fb2
c800551
eb03c6d
cb662f9
eb03c6d
cb662f9
 
a049b66
 
 
cb662f9
a049b66
cb662f9
 
 
 
 
 
 
 
 
 
 
a049b66
cb662f9
a049b66
cb662f9
c800551
a049b66
c800551
a049b66
 
 
1d33f2b
cb662f9
 
a049b66
cb662f9
a049b66
cb662f9
a049b66
cb662f9
a049b66
 
cb662f9
a049b66
 
 
 
 
 
 
 
cb662f9
 
 
 
a049b66
cb662f9
a049b66
 
cb662f9
 
 
a049b66
 
 
 
f733237
a049b66
cb662f9
a049b66
cb662f9
 
a049b66
cb662f9
a049b66
 
 
c800551
a049b66
 
c800551
 
cb662f9
a049b66
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f733237
a049b66
 
 
cb662f9
 
 
 
 
 
c800551
 
 
cb662f9
c800551
 
 
7d9aafd
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
import gradio as gr
import os
import shutil
import subprocess
from llama_cpp import Llama
from pymatgen.core import Structure
from huggingface_hub import hf_hub_download

# --------------------
# Paths & Config
# --------------------
GEN_DIR = "generation_cifs"

REPO_ID = "andreamena/crystext-gguf"
FILENAME = "crystext-rl.gguf"

# Fixed evaluation compositions (from your dataset)
KNOWN_MATERIALS = [
    "NaTiVS4",
    "ScAlRh2",
    "Fe3Te2Se",
    "Eu2CrSbO6",
    "Fe(HO)2",
    "Sm(CoB)2",
    "GaTe",
    "Zr2VRe3",
    "LiMnO2",
    "AlAu4",
]

# --------------------
# Load GGUF model
# --------------------
print("⏳ Downloading Model from Hub... this may take a minute...")
try:
    model_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
    print(f"✅ Model downloaded to: {model_path}")

    print("⏳ Initializing CPU Engine...")
    llm = Llama(
        model_path=model_path,
        n_ctx=2048,
        n_threads=2,      # HF free tier = 2 vCPUs
        verbose=True
    )
    print("✅ Engine Ready!")
except Exception as e:
    print(f"❌ Error setting up model: {e}")
    llm = None

# --------------------
# Generation helpers
# --------------------
def generate_single_cif(composition, space_group, temperature):
    prompt = f"""Below is an instruction that describes a task, paired with an input that provides further context.

### Instruction:
Generate CIF for the given material description

### Input:
Material composition is {composition}. It has a space group number {space_group}.

### Response:
"""
    output = llm(
        prompt,
        max_tokens=2048,
        temperature=temperature,
        stop=["</s>"],
        echo=False
    )
    return output["choices"][0]["text"]

# --------------------
# Main pipeline
# --------------------
def run_demo(composition, space_group, num_samples, temperature, progress=gr.Progress()):
    shutil.rmtree(GEN_DIR, ignore_errors=True)
    os.makedirs(GEN_DIR, exist_ok=True)

    if llm is None:
        return None, "Model failed to load.", "N/A"

    status_msg = f"Generating {num_samples} candidates (CPU: ~6 min/sample)\n"
    generated_files = []

    for i in progress.tqdm(range(int(num_samples)), desc="Generating structures"):
        try:
            cif_text = generate_single_cif(composition, space_group, temperature)
            cif_text = cif_text.replace("```cif", "").replace("```", "").strip()

            path = os.path.join(GEN_DIR, f"sample_{i}.cif")
            with open(path, "w") as f:
                f.write(cif_text)

            generated_files.append(path)
        except Exception as e:
            print(f"Sample {i} failed: {e}")

    if not generated_files:
        return None, "No valid CIFs generated.", "N/A"

    # Evaluation
    progress(0.9, desc="Evaluating structures...")
    try:
        subprocess.run(["python", "select_best_cif.py"], check=True)
        eval_text = subprocess.check_output(["python", "evaluation.py"], text=True)

        final_path = "final.cif"
        if os.path.exists(final_path):
            struct = Structure.from_file(final_path)
            sg_info = struct.get_space_group_info(symprec=0.1)
            return final_path, f"{status_msg}\n{eval_text}", f"{sg_info}"
        else:
            return None, f"{status_msg}\nEvaluation finished but no final.cif found.", "N/A"

    except Exception as e:
        return None, f"Evaluation error: {e}", "N/A"

# --------------------
# Gradio UI
# --------------------
with gr.Blocks() as demo:
    gr.Markdown("# CrysText-RL: Crystal Generation")
    gr.Markdown(f"Running GGUF model sourced from: `{REPO_ID}`")

    with gr.Row():
        with gr.Column():
            composition = gr.Dropdown(
                choices=KNOWN_MATERIALS,
                value="NaTiVS4",
                label="Composition (evaluation set)"
            )

            space_group = gr.Slider(
                minimum=1,
                maximum=230,
                step=1,
                value=8,
                label="Target Space Group"
            )

            num_samples = gr.Slider(
                1, 6, value=1, step=1,
                label="Samples (CPU-limited)"
            )

            temperature = gr.Slider(
                0.6, 1.0, value=0.8, step=0.05,
                label="Temperature"
            )

            run_btn = gr.Button("Generate Crystal Structure", variant="primary")

        with gr.Column():
            sg_output = gr.Textbox(label="Predicted Space Group")
            eval_output = gr.Textbox(label="Evaluation Metrics", lines=10)
            best_cif = gr.File(label="Download Best CIF")

    run_btn.click(
        run_demo,
        inputs=[composition, space_group, num_samples, temperature],
        outputs=[best_cif, eval_output, sg_output],
    )

demo.launch(
    auth=(os.environ.get("app_user"), os.environ.get("app_pass")),
    ssr_mode=False

)