upload main file

app.py

@@ -0,0 +1,304 @@
+import torch
+import torch.nn as nn
+import gradio as gr
+import os
+import pickle
+
+class LayerNorm(nn.Module):
+    def __init__(self, emb_dim):
+        super().__init__()
+        self.eps = 1e-5
+        self.scale = nn.Parameter(torch.ones(emb_dim))
+        self.shift = nn.Parameter(torch.zeros(emb_dim))
+
+    def forward(self, x):
+        mean = x.mean(dim=-1, keepdim=True)
+        var = x.var(dim=-1, keepdim=True)
+        norm_x = (x - mean) / torch.sqrt(var + self.eps)
+        return self.scale * norm_x + self.shift
+
+
+class GELU(nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    def forward(self, x):
+        return 0.5 * x * (1 + torch.tanh(torch.sqrt(torch.tensor(2.0 / torch.pi)) * (x + 0.044715 * torch.pow(x, 3))))
+
+
+class MultiHeadAttention(nn.Module):
+    def __init__(self, d_in, d_out, context_length, dropout, num_head, qkv_bias=False):
+        super().__init__()
+        assert (d_out % num_head == 0)
+
+        self.d_out = d_out
+        self.num_head = num_head
+        self.head_dim = d_out // num_head
+
+        self.W_query = torch.nn.Linear(d_in, d_out, bias=qkv_bias)
+        self.W_key = torch.nn.Linear(d_in, d_out, bias=qkv_bias)
+        self.W_value = torch.nn.Linear(d_in, d_out, bias=qkv_bias)
+
+        self.out_proj = torch.nn.Linear(d_out, d_out)
+        self.dropout = torch.nn.Dropout(dropout)
+        self.register_buffer("mask", torch.triu(torch.ones(context_length, context_length), diagonal=1))
+
+    def forward(self, x):
+        b, num_tokens, d_in = x.shape
+
+        keys = self.W_key(x)
+        queries = self.W_query(x)
+        values = self.W_value(x)
+
+        keys = keys.view(b, num_tokens, self.num_head, self.head_dim)
+        values = values.view(b, num_tokens, self.num_head, self.head_dim)
+        queries = queries.view(b, num_tokens, self.num_head, self.head_dim)
+
+        keys = keys.transpose(1, 2)
+        values = values.transpose(1, 2)
+        queries = queries.transpose(1, 2)
+
+        attn_score = queries @ keys.transpose(2, 3)
+
+        mask_bool = self.mask.to(torch.bool)[:num_tokens, :num_tokens]
+
+        attn_score.masked_fill_(mask_bool, -torch.inf)
+
+        attn_weight = torch.softmax(attn_score / keys.shape[-1] ** 0.5, dim=-1)
+        attn_weight = self.dropout(attn_weight)
+
+        context_vector = (attn_weight @ values).transpose(1, 2)
+
+        context_vector = context_vector.contiguous().view(b, num_tokens, self.d_out)
+        context_vector = self.out_proj(context_vector)
+
+        return context_vector
+
+
+class FeedForward(nn.Module):
+    def __init__(self, cfg):
+        super().__init__()
+        self.layers = nn.Sequential(
+            nn.Linear(cfg["emb_dim"], 4 * cfg["emb_dim"]),
+            GELU(),
+            nn.Linear(4 * cfg["emb_dim"], cfg["emb_dim"])
+        )
+
+    def forward(self, x):
+        return self.layers(x)
+
+
+class TransformerBlock(nn.Module):
+    def __init__(self, cfg):
+        super().__init__()
+        self.att = MultiHeadAttention(
+            d_in=cfg["emb_dim"],
+            d_out=cfg["emb_dim"],
+            context_length=cfg["context_length"],
+            num_head=cfg["n_heads"],
+            dropout=cfg.get("drop_rate", 0.0),
+            qkv_bias=cfg.get("qkv_bias", False)
+        )
+
+        self.ff = FeedForward(cfg)
+        self.norm1 = LayerNorm(cfg["emb_dim"])
+        self.norm2 = LayerNorm(cfg["emb_dim"])
+        self.drop_shortcut = nn.Dropout(cfg.get("drop_rate", 0.0))
+
+    def forward(self, x):
+        shortcut = x
+        x = self.norm1(x)
+        x = self.att(x)
+        x = self.drop_shortcut(x)
+        x = x + shortcut
+
+        shortcut = x
+        x = self.norm2(x)
+        x = self.ff(x)
+        x = self.drop_shortcut(x)
+        x = x + shortcut
+        return x
+
+
+class GPTModel(nn.Module):
+    def __init__(self, cfg):
+        super().__init__()
+        self.tok_emb = nn.Embedding(cfg["vocab_size"], cfg["emb_dim"])
+        self.pos_emb = nn.Embedding(cfg["context_length"], cfg["emb_dim"])
+        self.drop_emb = nn.Dropout(cfg.get("drop_rate", 0.0))
+
+        self.trf_blocks = nn.Sequential(
+            *[TransformerBlock(cfg) for _ in range(cfg["n_layers"]) ]
+        )
+
+        self.final_norm = LayerNorm(cfg["emb_dim"])
+        self.out_head = nn.Linear(cfg["emb_dim"], cfg["vocab_size"], bias=False)
+
+    def forward(self, in_idx):
+        batch_size, seq_len = in_idx.shape
+        tok_embeds = self.tok_emb(in_idx)
+        pos_embeds = self.pos_emb(torch.arange(seq_len, device=in_idx.device))
+        x = tok_embeds + pos_embeds
+        x = self.drop_emb(x)
+        x = self.trf_blocks(x)
+        x = self.final_norm(x)
+        logits = self.out_head(x)
+        return logits
+
+
+model_path = "review_classifier_model.pth"
+if not os.path.exists(model_path):
+    raise FileNotFoundError(f"{model_path} not found. Please check the path.")
+
+try:
+    loaded_full = None
+    safe_ctx = getattr(torch.serialization, "safe_globals", None)
+    add_safe = getattr(torch.serialization, "add_safe_globals", None)
+
+    if safe_ctx is not None:
+        try:
+            with torch.serialization.safe_globals([GPTModel]):
+                loaded_full = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+        except Exception:
+            loaded_full = None
+    elif add_safe is not None:
+        try:
+            # older helper: register globally then load
+            torch.serialization.add_safe_globals([GPTModel])
+            loaded_full = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+        except Exception:
+            loaded_full = None
+    else:
+        # If neither helper exists, try loading with weights_only=False (may execute code during unpickle).
+        try:
+            loaded_full = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+        except Exception:
+            loaded_full = None
+
+    if loaded_full is not None and hasattr(loaded_full, "state_dict") and not isinstance(loaded_full, dict):
+        model = loaded_full
+        print(f"Loaded full model object from {model_path}")
+    else:
+        state = None
+        try:
+            state = torch.load(model_path, map_location=torch.device("cpu"), weights_only=True)
+        except Exception:
+            try:
+                if safe_ctx is not None:
+                    with torch.serialization.safe_globals([GPTModel]):
+                        tmp = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+                elif add_safe is not None:
+                    torch.serialization.add_safe_globals([GPTModel])
+                    tmp = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+                else:
+                    tmp = torch.load(model_path, map_location=torch.device("cpu"), weights_only=False)
+
+                if hasattr(tmp, "state_dict"):
+                    state = tmp.state_dict()
+                else:
+                    state = tmp
+            except Exception as e:
+                raise RuntimeError(f"Unable to load checkpoint as full model or weights-only. Last error: {e}")
+
+        if isinstance(state, dict):
+            print("Attempting to load checkpoint state into a GPTModel instance...")
+            BASE_CONFIG = {
+                "vocab_size": 50257,
+                "context_length": 1024,
+                "drop_rate": 0.0,
+                "qkv_bias": True,
+                "emb_dim": 768,
+                "n_layers": 12,
+                "n_heads": 12,
+            }
+            model = GPTModel(BASE_CONFIG)
+
+            if "model_state_dict" in state:
+                state_dict = state["model_state_dict"]
+            elif "state_dict" in state:
+                state_dict = state["state_dict"]
+            else:
+                state_dict = state
+
+            model.load_state_dict(state_dict, strict=False)
+            print("Loaded state_dict into GPTModel instance (non-strict).")
+        else:
+            raise RuntimeError("Unrecognized checkpoint format and unable to construct model from checkpoint.")
+except Exception as e:
+    raise RuntimeError(f"Failed to load model checkpoint: {e}")
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model.to(device)
+model.eval()
+print(f"Model loaded and moved to {device}")
+
+tokenizer_path = "tokenizer.pkl"
+if os.path.exists(tokenizer_path):
+    with open(tokenizer_path, "rb") as f:
+        tokenizer = pickle.load(f)
+    print(f"Tokenizer loaded from {tokenizer_path}")
+else:
+    raise FileNotFoundError(f"{tokenizer_path} not found. Please check the path.")
+
+MAX_SEQUENCE_LENGTH = 120
+
+def classify_review(text, model, tokenizer_obj, device, max_length=MAX_SEQUENCE_LENGTH, pad_token_id=50256):
+    model.eval()
+    input_ids = tokenizer_obj.encode(text)
+    input_ids = input_ids[:max_length] + [pad_token_id] * (max_length - len(input_ids))
+    input_tensor = torch.tensor(input_ids, device=device).unsqueeze(0)
+    with torch.no_grad():
+        logits = model(input_tensor)[:, -1, :]
+    predicted_label = torch.argmax(logits, dim=-1).item()
+    return "spam" if predicted_label == 1 else "not spam"
+
+def chatbot_classify(message, history):
+    result = classify_review(
+        message,
+        model,
+        tokenizer,
+        device,
+        max_length=MAX_SEQUENCE_LENGTH
+    )
+    return result
+
+print("Launching Gradio interface...")
+
+iface = gr.ChatInterface(
+    chatbot_classify,
+    title="📬 Spam Detection System",
+    description="Enter an SMS message below...",
+    theme="compact",  
+    css="""
+        /* Customize chat bubble colors */
+        .chatbot-message {
+            background-color: #e0f7fa; /* Light cyan */
+            color: #006064;           /* Dark teal text */
+            font-weight: 600;
+            border-radius: 12px;
+            padding: 12px;
+        }
+        .user-message {
+            background-color: #c8e6c9; /* Light green */
+            color: #1b5e20;            /* Dark green text */
+            font-weight: 600;
+            border-radius: 12px;
+            padding: 12px;
+        }
+        .chat-ending-message {
+            font-style: italic;
+            color: #555;
+        }
+    """,
+)
+
+ICON_CDN = "https://img.icons8.com/color/48/mail-envelope.png"
+
+custom_head_html = f"""
+    <link rel="icon" href="{ICON_CDN}" type="image/x-icon">
+"""
+
+iface.launch(
+    share=True,
+    favicon_path=ICON_CDN
+)