"""
SmolLM2-135M Text Generation - Gradio App
Trained from scratch on Suits TV series scripts
"""

import torch
import torch.nn as nn
from torch.nn import functional as F
from dataclasses import dataclass
from typing import Optional, Tuple
import tiktoken
import gradio as gr

# ============================================================================
# Model Architecture
# ============================================================================

@dataclass
class SmolLM2Config:
    """SmolLM2-135M Configuration"""
    vocab_size: int = 50304
    hidden_size: int = 576
    intermediate_size: int = 1536
    num_hidden_layers: int = 30
    num_attention_heads: int = 9
    num_key_value_heads: int = 3
    max_position_embeddings: int = 2048
    rms_norm_eps: float = 1e-5
    rope_theta: float = 10000.0
    hidden_act: str = "silu"
    initializer_range: float = 0.041666666666666664
    tie_word_embeddings: bool = True
    bos_token_id: int = 0
    eos_token_id: int = 0

    @property
    def head_dim(self) -> int:
        return self.hidden_size // self.num_attention_heads


class RMSNorm(nn.Module):
    """Root Mean Square Layer Normalization"""
    def __init__(self, hidden_size: int, eps: float = 1e-5):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.eps = eps

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        rms = torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
        return x * rms * self.weight


class RotaryEmbedding(nn.Module):
    """Rotary Position Embedding (RoPE)"""
    def __init__(self, dim: int, max_position_embeddings: int = 2048, theta: float = 10000.0):
        super().__init__()
        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.theta = theta

        inv_freq = 1.0 / (self.theta ** (torch.arange(0, self.dim, 2, dtype=torch.float32) / self.dim))
        self.register_buffer("inv_freq", inv_freq, persistent=False)
        self._set_cos_sin_cache(max_position_embeddings)

    def _set_cos_sin_cache(self, seq_len: int):
        self.max_seq_len_cached = seq_len
        t = torch.arange(seq_len, dtype=torch.float32)
        freqs = torch.outer(t, self.inv_freq)
        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", emb.cos(), persistent=False)
        self.register_buffer("sin_cached", emb.sin(), persistent=False)

    def forward(self, x: torch.Tensor, seq_len: int) -> Tuple[torch.Tensor, torch.Tensor]:
        if seq_len > self.max_seq_len_cached:
            self._set_cos_sin_cache(seq_len)
        return (
            self.cos_cached[:seq_len].to(x.dtype),
            self.sin_cached[:seq_len].to(x.dtype)
        )


def rotate_half(x: torch.Tensor) -> torch.Tensor:
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)


def apply_rotary_pos_emb(q: torch.Tensor, k: torch.Tensor, cos: torch.Tensor, sin: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
    cos = cos.unsqueeze(0).unsqueeze(0)
    sin = sin.unsqueeze(0).unsqueeze(0)
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed


class GroupedQueryAttention(nn.Module):
    """Grouped Query Attention (GQA)"""
    def __init__(self, config: SmolLM2Config):
        super().__init__()
        self.config = config
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.num_kv_heads = config.num_key_value_heads
        self.head_dim = config.head_dim
        self.num_kv_groups = self.num_heads // self.num_kv_heads

        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
        self.k_proj = nn.Linear(self.hidden_size, self.num_kv_heads * self.head_dim, bias=False)
        self.v_proj = nn.Linear(self.hidden_size, self.num_kv_heads * self.head_dim, bias=False)
        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)

        self.rotary_emb = RotaryEmbedding(
            self.head_dim,
            max_position_embeddings=config.max_position_embeddings,
            theta=config.rope_theta
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, T, C = x.size()

        q = self.q_proj(x)
        k = self.k_proj(x)
        v = self.v_proj(x)

        q = q.view(B, T, self.num_heads, self.head_dim).transpose(1, 2)
        k = k.view(B, T, self.num_kv_heads, self.head_dim).transpose(1, 2)
        v = v.view(B, T, self.num_kv_heads, self.head_dim).transpose(1, 2)

        cos, sin = self.rotary_emb(x, T)
        q, k = apply_rotary_pos_emb(q, k, cos, sin)

        k = k.repeat_interleave(self.num_kv_groups, dim=1)
        v = v.repeat_interleave(self.num_kv_groups, dim=1)

        y = F.scaled_dot_product_attention(q, k, v, is_causal=True)
        y = y.transpose(1, 2).contiguous().view(B, T, self.hidden_size)
        y = self.o_proj(y)
        return y


class SwiGLUMLP(nn.Module):
    """SwiGLU Feed-Forward Network"""
    def __init__(self, config: SmolLM2Config):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.intermediate_size = config.intermediate_size

        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
        self.act_fn = nn.SiLU()

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))


class SmolLM2Block(nn.Module):
    """SmolLM2 Transformer Block"""
    def __init__(self, config: SmolLM2Config):
        super().__init__()
        self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.self_attn = GroupedQueryAttention(config)
        self.post_attention_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.mlp = SwiGLUMLP(config)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = x + self.self_attn(self.input_layernorm(x))
        x = x + self.mlp(self.post_attention_layernorm(x))
        return x


class SmolLM2(nn.Module):
    """SmolLM2-135M Model"""
    def __init__(self, config: SmolLM2Config):
        super().__init__()
        self.config = config

        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
        self.layers = nn.ModuleList([SmolLM2Block(config) for _ in range(config.num_hidden_layers)])
        self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

        if config.tie_word_embeddings:
            self.lm_head.weight = self.embed_tokens.weight

    def forward(self, input_ids: torch.Tensor, targets: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
        B, T = input_ids.size()
        x = self.embed_tokens(input_ids)

        for layer in self.layers:
            x = layer(x)

        x = self.norm(x)
        logits = self.lm_head(x)

        loss = None
        if targets is not None:
            loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1))

        return logits, loss


# ============================================================================
# Load Model
# ============================================================================

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

# Load model
config = SmolLM2Config()
model = SmolLM2(config)

# Load trained weights
checkpoint = torch.load("smollm2_135m_final.pt", map_location=device, weights_only=False)
model.load_state_dict(checkpoint['model_state_dict'])
model.to(device)
model.eval()

print(f"Model loaded successfully! Parameters: {sum(p.numel() for p in model.parameters()):,}")

# Load tokenizer
tokenizer = tiktoken.get_encoding('gpt2')


# ============================================================================
# Generation Function
# ============================================================================

def generate_text(
    prompt: str,
    max_new_tokens: int = 100,
    temperature: float = 0.8,
    top_k: int = 50,
    top_p: float = 0.9,
) -> str:
    """Generate text from a prompt"""
    if not prompt.strip():
        return "Please enter a prompt."

    model.eval()

    # Encode prompt
    tokens = tokenizer.encode(prompt)
    tokens = torch.tensor(tokens, dtype=torch.long, device=device).unsqueeze(0)

    # Generate tokens
    with torch.no_grad():
        for _ in range(max_new_tokens):
            # Crop to max position embeddings
            idx_cond = tokens[:, -config.max_position_embeddings:]

            # Get predictions
            logits, _ = model(idx_cond)
            logits = logits[:, -1, :] / temperature

            # Top-k filtering
            if top_k is not None and top_k > 0:
                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                logits[logits < v[:, [-1]]] = float('-inf')

            # Top-p (nucleus) filtering
            if top_p is not None and top_p < 1.0:
                sorted_logits, sorted_indices = torch.sort(logits, descending=True)
                cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
                sorted_indices_to_remove = cumulative_probs > top_p
                sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
                sorted_indices_to_remove[..., 0] = 0
                indices_to_remove = sorted_indices_to_remove.scatter(1, sorted_indices, sorted_indices_to_remove)
                logits[indices_to_remove] = float('-inf')

            # Sample
            probs = F.softmax(logits, dim=-1)
            next_token = torch.multinomial(probs, num_samples=1)
            tokens = torch.cat([tokens, next_token], dim=1)

    # Decode
    generated = tokenizer.decode(tokens[0].tolist())
    return generated


# ============================================================================
# Gradio Interface
# ============================================================================

title = "SmolLM2-135M Text Generator"
description = """
## About This Model

This is a **SmolLM2-135M** model trained from scratch on dialogue scripts from the TV series "Suits".

### Model Architecture
- **Type**: Llama-based decoder-only transformer
- **Parameters**: ~135M
- **Features**: RMSNorm, RoPE, Grouped Query Attention (GQA), SwiGLU MLP

### Training Details
- Trained for 5,050 steps
- Sequence length: 1024 tokens
- Uses GPT-2 tokenizer

Enter a prompt below and adjust the generation parameters to see what the model generates!
"""

examples = [
    ["Harvey walked into the office and said,"],
    ["The legal case was complicated because"],
    ["Once upon a time"],
    ["In a world where lawyers"],
    ["Mike looked at the contract and noticed"],
]

# Create interface
with gr.Blocks(title=title, theme=gr.themes.Soft()) as demo:
    gr.Markdown(f"# {title}")
    gr.Markdown(description)

    with gr.Row():
        with gr.Column(scale=2):
            prompt_input = gr.Textbox(
                label="Prompt",
                placeholder="Enter your prompt here...",
                lines=3,
            )

            with gr.Row():
                max_tokens_slider = gr.Slider(
                    minimum=10,
                    maximum=500,
                    value=100,
                    step=10,
                    label="Max New Tokens",
                )
                temperature_slider = gr.Slider(
                    minimum=0.1,
                    maximum=2.0,
                    value=0.8,
                    step=0.1,
                    label="Temperature",
                )

            with gr.Row():
                top_k_slider = gr.Slider(
                    minimum=1,
                    maximum=100,
                    value=50,
                    step=1,
                    label="Top-K",
                )
                top_p_slider = gr.Slider(
                    minimum=0.1,
                    maximum=1.0,
                    value=0.9,
                    step=0.05,
                    label="Top-P (Nucleus)",
                )

            generate_btn = gr.Button("Generate", variant="primary")

        with gr.Column(scale=2):
            output_text = gr.Textbox(
                label="Generated Text",
                lines=15,
            )

    gr.Markdown("### Example Prompts")
    gr.Examples(
        examples=examples,
        inputs=prompt_input,
    )

    # Connect the generate button
    generate_btn.click(
        fn=generate_text,
        inputs=[prompt_input, max_tokens_slider, temperature_slider, top_k_slider, top_p_slider],
        outputs=output_text,
    )

    # Also generate on Enter key
    prompt_input.submit(
        fn=generate_text,
        inputs=[prompt_input, max_tokens_slider, temperature_slider, top_k_slider, top_p_slider],
        outputs=output_text,
    )

    gr.Markdown("""
    ---
    ### Parameter Guide
    - **Temperature**: Higher = more creative/random, Lower = more focused/deterministic
    - **Top-K**: Only sample from the top K most likely tokens
    - **Top-P**: Only sample from tokens whose cumulative probability is below P
    - **Max New Tokens**: Maximum number of tokens to generate

    ---
    *Model trained from scratch using PyTorch. Architecture based on SmolLM2-135M (Llama-style).*
    """)


if __name__ == "__main__":
    demo.launch()