File size: 8,611 Bytes
aee08b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
"""
Genomic Semantic Search API with FastAPI
=========================================
Search genomic sequences using your pre-trained transformer embeddings.
"""

import pickle
from pathlib import Path
from typing import Optional
import numpy as np
import pandas as pd
from fastapi import FastAPI, HTTPException
from fastapi.staticfiles import StaticFiles
from fastapi.responses import HTMLResponse, FileResponse
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
import faiss
import torch
import torch.nn as nn
from x_transformers import TransformerWrapper, Encoder
import tiktoken

# ============================================================================
# Configuration
# ============================================================================

DATA_DIR = Path("data")
INDEX_PATH = DATA_DIR / "data/faiss.index"
METADATA_PATH = DATA_DIR / "data/metadata.pkl"
EMBEDDINGS_PATH = DATA_DIR / "data/embeddings.npy"

# Model paths - update these to your actual paths
MODEL_WEIGHTS_PATH = DATA_DIR / "data/bpe_plus_special_tokens_model.pt"
TOKENIZER_PATH = DATA_DIR / "data/bpe_plus_special_tokens_tokenizer.json"

# ============================================================================
# Model Definition
# ============================================================================

class GenomicTransformer(nn.Module):
    def __init__(self, vocab_size=40000, hidden_dim=512, layers=12, heads=8, max_length=6000):
        super().__init__()
        self.model = TransformerWrapper(
            num_tokens=vocab_size,
            max_seq_len=max_length,
            attn_layers=Encoder(
                dim=hidden_dim,
                depth=layers,
                heads=heads,
                rotary_pos_emb=True,
                attn_orthog_projected_values=True,
                attn_orthog_projected_values_per_head=True,
                attn_flash=True
            )
        )

    def forward(self, input_ids, return_embeddings=False):
        return self.model(input_ids, return_embeddings=return_embeddings)

# ============================================================================
# App Setup
# ============================================================================

app = FastAPI(
    title="Genomic Semantic Search",
    description="Search genomic sequences using transformer embeddings",
    version="1.0.0"
)

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Global state
device: torch.device = None
model: Optional[GenomicTransformer] = None
encoder: Optional[tiktoken.Encoding] = None
index: Optional[faiss.IndexFlatIP] = None
metadata: Optional[pd.DataFrame] = None

# ============================================================================
# Models
# ============================================================================

class SearchRequest(BaseModel):
    query: str  # The genomic sequence to search for
    top_k: int = 10

class SearchResult(BaseModel):
    rank: int
    score: float
    sequence: str
    metadata: dict

class SearchResponse(BaseModel):
    query: str
    results: list[SearchResult]
    total_indexed: int

class IndexStats(BaseModel):
    total_documents: int
    embedding_dimension: int
    model_name: str
    device: str

# ============================================================================
# Startup
# ============================================================================

@app.on_event("startup")
async def startup():
    """Load the model, tokenizer, and FAISS index on startup."""
    global device, model, encoder, index, metadata

    # Setup device
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    print(f"Using device: {device}")

    # Load tokenizer
    print("Loading tokenizer...")
    if TOKENIZER_PATH.exists():
        with open(TOKENIZER_PATH, "rb") as f:
            tokenizer_data = pickle.load(f)
        encoder = tiktoken.Encoding(
            name="genomic_bpe",
            pat_str=tokenizer_data['pattern'],
            mergeable_ranks=tokenizer_data['mergable_ranks'],
            special_tokens={}
        )
        print("Tokenizer loaded successfully")
    else:
        print(f"WARNING: Tokenizer not found at {TOKENIZER_PATH}")

    # Load model
    print("Loading model...")
    if MODEL_WEIGHTS_PATH.exists():
        model = GenomicTransformer(
            vocab_size=40_000, hidden_dim=512, layers=12, heads=8
        )
        weights = torch.load(MODEL_WEIGHTS_PATH, map_location=device)
        model.load_state_dict(weights)
        model = model.to(device)
        model.eval()
        print("Model loaded successfully")
    else:
        print(f"WARNING: Model weights not found at {MODEL_WEIGHTS_PATH}")

    # Load FAISS index
    if INDEX_PATH.exists() and METADATA_PATH.exists():
        print("Loading FAISS index...")
        index = faiss.read_index(str(INDEX_PATH))
        with open(METADATA_PATH, "rb") as f:
            metadata = pickle.load(f)
        print(f"Index loaded with {index.ntotal} documents")
    else:
        print(f"WARNING: Index not found at {INDEX_PATH}")

# ============================================================================
# API Endpoints
# ============================================================================

@app.get("/", response_class=HTMLResponse)
async def root():
    """Serve the search frontend."""
    return FileResponse("index.html")

@app.get("/api/health")
async def health():
    """Health check endpoint."""
    return {
        "status": "healthy",
        "model_loaded": model is not None,
        "index_loaded": index is not None,
        "tokenizer_loaded": encoder is not None,
        "device": str(device)
    }

@app.get("/api/stats", response_model=IndexStats)
async def get_stats():
    """Get statistics about the current index."""
    if index is None:
        raise HTTPException(status_code=404, detail="No index loaded")

    return IndexStats(
        total_documents=index.ntotal,
        embedding_dimension=index.d,
        model_name="GenomicTransformer (512d, 12 layers)",
        device=str(device)
    )

@app.post("/api/search", response_model=SearchResponse)
async def search(request: SearchRequest):
    """
    Perform semantic search over genomic sequences.

    - **query**: The genomic sequence to search for (e.g., "ATCGATCG...")
    - **top_k**: Number of results to return (default: 10)
    """
    if index is None or metadata is None:
        raise HTTPException(status_code=404, detail="No index loaded")
    if model is None or encoder is None:
        raise HTTPException(status_code=503, detail="Model or tokenizer not loaded")
    if index.ntotal == 0:
        raise HTTPException(status_code=404, detail="Index is empty")

    # Encode the query sequence
    try:
        encodings = encoder.encode_ordinary(request.query)
        query_tensor = torch.tensor([encodings]).long().to(device)

        with torch.no_grad():
            query_embedding = model(query_tensor, return_embeddings=True)
            query_embedding = query_embedding.mean(dim=1).cpu().numpy()

        query_embedding = query_embedding.astype(np.float32)
    except Exception as e:
        raise HTTPException(status_code=400, detail=f"Failed to encode query: {str(e)}")

    # Search
    k = min(request.top_k, index.ntotal)
    scores, indices = index.search(query_embedding, k)

    # Build results
    results = []
    for rank, (score, idx) in enumerate(zip(scores[0], indices[0]), 1):
        if idx == -1:
            continue

        row = metadata.iloc[idx]
        meta_dict = row.to_dict()
        sequence = meta_dict.pop("__sequence__", "")

        results.append(SearchResult(
            rank=rank,
            score=float(score),
            sequence=sequence,
            metadata=meta_dict
        ))

    return SearchResponse(
        query=request.query[:100] + "..." if len(request.query) > 100 else request.query,
        results=results,
        total_indexed=index.ntotal
    )

@app.get("/api/sample")
async def get_sample(n: int = 5):
    """Get a sample of indexed documents."""
    if metadata is None:
        raise HTTPException(status_code=404, detail="No index loaded")

    sample = metadata.head(n)
    return {
        "total": len(metadata),
        "sample": sample.to_dict(orient="records")
    }

# Mount  files
# app.mount("/static", StaticFiles(directory="static"), name="static")

if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8080)