"""

"""
from __future__ import annotations

import os
from dataclasses import dataclass

import gradio as gr
import tiktoken
import torch
import torch.nn as nn
import torch.nn.functional as F

CKPT_PATH = os.environ.get("CKPT_PATH", "fable5_transformer.pt")
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"


# ---------------------------------------------------------------------------
# Model (must match train_transformer.py)
# ---------------------------------------------------------------------------
@dataclass
class Config:
    vocab_size: int = 50257
    n_layer: int = 6
    n_head: int = 8
    n_embd: int = 256
    block_size: int = 256
    dropout: float = 0.0


class CausalSelfAttention(nn.Module):
    def __init__(self, cfg: Config):
        super().__init__()
        self.n_head = cfg.n_head
        self.n_embd = cfg.n_embd
        self.qkv = nn.Linear(cfg.n_embd, 3 * cfg.n_embd, bias=False)
        self.proj = nn.Linear(cfg.n_embd, cfg.n_embd, bias=False)
        self.drop = nn.Dropout(cfg.dropout)

    def forward(self, x):
        B, T, C = x.shape
        q, k, v = self.qkv(x).split(self.n_embd, dim=2)
        hd = C // self.n_head
        q = q.view(B, T, self.n_head, hd).transpose(1, 2)
        k = k.view(B, T, self.n_head, hd).transpose(1, 2)
        v = v.view(B, T, self.n_head, hd).transpose(1, 2)
        y = F.scaled_dot_product_attention(q, k, v, is_causal=True)
        y = y.transpose(1, 2).contiguous().view(B, T, C)
        return self.drop(self.proj(y))


class MLP(nn.Module):
    def __init__(self, cfg: Config):
        super().__init__()
        self.fc = nn.Linear(cfg.n_embd, 4 * cfg.n_embd, bias=False)
        self.proj = nn.Linear(4 * cfg.n_embd, cfg.n_embd, bias=False)
        self.drop = nn.Dropout(cfg.dropout)

    def forward(self, x):
        return self.drop(self.proj(F.gelu(self.fc(x))))


class Block(nn.Module):
    def __init__(self, cfg: Config):
        super().__init__()
        self.ln1 = nn.LayerNorm(cfg.n_embd)
        self.attn = CausalSelfAttention(cfg)
        self.ln2 = nn.LayerNorm(cfg.n_embd)
        self.mlp = MLP(cfg)

    def forward(self, x):
        x = x + self.attn(self.ln1(x))
        x = x + self.mlp(self.ln2(x))
        return x


class TinyGPT(nn.Module):
    def __init__(self, cfg: Config):
        super().__init__()
        self.cfg = cfg
        self.tok_emb = nn.Embedding(cfg.vocab_size, cfg.n_embd)
        self.pos_emb = nn.Embedding(cfg.block_size, cfg.n_embd)
        self.drop = nn.Dropout(cfg.dropout)
        self.blocks = nn.ModuleList([Block(cfg) for _ in range(cfg.n_layer)])
        self.ln_f = nn.LayerNorm(cfg.n_embd)
        self.head = nn.Linear(cfg.n_embd, cfg.vocab_size, bias=False)
        self.head.weight = self.tok_emb.weight

    def forward(self, idx):
        B, T = idx.shape
        pos = torch.arange(T, device=idx.device)
        x = self.drop(self.tok_emb(idx) + self.pos_emb(pos))
        for b in self.blocks:
            x = b(x)
        return self.head(self.ln_f(x))

    @torch.no_grad()
    def generate(self, idx, max_new=100, temperature=0.9, top_k=50):
        for _ in range(max_new):
            idx_cond = idx[:, -self.cfg.block_size :]
            logits = self(idx_cond)[:, -1, :] / max(temperature, 1e-5)
            if top_k:
                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                logits[logits < v[:, [-1]]] = -float("inf")
            probs = F.softmax(logits, dim=-1)
            nxt = torch.multinomial(probs, 1)
            idx = torch.cat([idx, nxt], dim=1)
        return idx


# ---------------------------------------------------------------------------
# Load
# ---------------------------------------------------------------------------
print(f"[load] device={DEVICE}  ckpt={CKPT_PATH}")
ckpt = torch.load(CKPT_PATH, map_location=DEVICE, weights_only=False)
cfg_dict = ckpt.get("cfg", {})
cfg = Config(**{k: v for k, v in cfg_dict.items() if k in Config.__dataclass_fields__})
cfg.dropout = 0.0

model = TinyGPT(cfg).to(DEVICE)
model.load_state_dict(ckpt["model"])
model.eval()
enc = tiktoken.get_encoding("gpt2")
print(f"[load] params={sum(p.numel() for p in model.parameters())/1e6:.2f}M")


# ---------------------------------------------------------------------------
# Inference
# ---------------------------------------------------------------------------
def generate(prompt: str, max_new_tokens: int, temperature: float, top_k: int, seed: int):
    if seed >= 0:
        torch.manual_seed(seed)
    text = prompt if prompt.strip() else "<|endoftext|>"
    ids = enc.encode(text, allowed_special={"<|endoftext|>"})
    if not ids:
        ids = [enc.eot_token]
    x = torch.tensor([ids], dtype=torch.long, device=DEVICE)
    out = model.generate(x, max_new=int(max_new_tokens), temperature=float(temperature), top_k=int(top_k))
    return enc.decode(out[0].tolist())


EXAMPLES = [
    ["USER: Make a new one, it should be", 120, 0.9, 50, -1],
    ["<|user|>\nWrite a bash script that", 120, 0.8, 40, 42],
    ["<|endoftext|>", 150, 1.0, 50, -1],
]

with gr.Blocks(title="Tiny Transformer") as demo:
    gr.Markdown(
        "# HamerLM\n"
        
    )
    with gr.Row():
        with gr.Column():
            prompt = gr.Textbox(label="Prompt", value="USER: Make a new one, it should be", lines=4)
            max_new = gr.Slider(16, 512, value=120, step=8, label="Max new tokens")
            temperature = gr.Slider(0.1, 1.5, value=0.9, step=0.05, label="Temperature")
            top_k = gr.Slider(1, 200, value=50, step=1, label="Top-k")
            seed = gr.Number(value=-1, precision=0, label="Seed (-1 = random)")
            btn = gr.Button("Generate", variant="primary")
        with gr.Column():
            out = gr.Textbox(label="Output", lines=18)

    btn.click(generate, [prompt, max_new, temperature, top_k, seed], out)
    gr.Examples(EXAMPLES, [prompt, max_new, temperature, top_k, seed])

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