import json, os
import torch, torch.nn as nn, torch.nn.functional as F
from tokenizers import Tokenizer


class Block(nn.Module):
    def __init__(self, d, n_head):
        super().__init__()
        self.n_head = n_head
        self.ln1, self.ln2 = nn.LayerNorm(d), nn.LayerNorm(d)
        self.qkv = nn.Linear(d, 3 * d, bias=False)
        self.proj = nn.Linear(d, d, bias=False)
        self.fc = nn.Linear(d, 4 * d, bias=False)
        self.fc2 = nn.Linear(4 * d, d, bias=False)

    def forward(self, x):
        B, T, C = x.shape
        q, k, v = self.qkv(self.ln1(x)).split(C, dim=2)
        q, k, v = [t.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) for t in (q, k, v)]
        y = F.scaled_dot_product_attention(q, k, v, is_causal=True)
        x = x + self.proj(y.transpose(1, 2).reshape(B, T, C))
        return x + self.fc2(F.gelu(self.fc(self.ln2(x))))


class GPT(nn.Module):
    def __init__(self, cfg):
        super().__init__()
        d = cfg["n_embd"]
        self.ctx = cfg["ctx"]
        self.tok_emb = nn.Embedding(cfg["vocab"], d)
        self.pos_emb = nn.Embedding(cfg["ctx"], d)
        self.blocks = nn.ModuleList([Block(d, cfg["n_head"]) for _ in range(cfg["n_layer"])])
        self.ln_f = nn.LayerNorm(d)
        self.head = nn.Linear(d, cfg["vocab"], bias=False)
        self.head.weight = self.tok_emb.weight

    def forward(self, idx):
        x = self.tok_emb(idx) + self.pos_emb(torch.arange(idx.shape[1], device=idx.device))
        for blk in self.blocks:
            x = blk(x)
        return self.head(self.ln_f(x))

    @torch.no_grad()
    def generate(self, idx, max_new=100, temperature=0.7, top_k=40, stop_id=None):
        for _ in range(max_new):
            logits = self(idx[:, -self.ctx:])[:, -1, :] / max(temperature, 1e-5)
            if top_k:
                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                logits[logits < v[:, [-1]]] = -float("inf")
            nxt = torch.multinomial(F.softmax(logits, dim=-1), 1)
            idx = torch.cat([idx, nxt], dim=1)
            if stop_id is not None and nxt.item() == stop_id:
                break
        return idx


def load(path, device="cpu"):
    with open(os.path.join(path, "config.json")) as f:
        cfg = json.load(f)
    model = GPT(cfg)
    model.load_state_dict(torch.load(os.path.join(path, "model.pt"), map_location=device))
    tok = Tokenizer.from_file(os.path.join(path, "tokenizer.json"))
    return model.to(device).eval(), tok


def reply(model, tok, history, user_msg, temperature=0.7, top_k=40, max_new=120):
    # history is a list of (user, assistant) string pairs
    USER, ASSIST, END = [tok.token_to_id(t) for t in ("<|user|>", "<|assistant|>", "<|end|>")]
    ids = []
    for u, a in history:
        ids += [USER] + tok.encode(u).ids + [END, ASSIST] + tok.encode(a).ids + [END]
    ids += [USER] + tok.encode(user_msg).ids + [END, ASSIST]
    ids = ids[-(model.ctx - 1):]
    device = next(model.parameters()).device
    out = model.generate(torch.tensor([ids], device=device), max_new=max_new,
                         temperature=temperature, top_k=top_k, stop_id=END)
    new = out[0, len(ids):].tolist()
    if new and new[-1] == END:
        new = new[:-1]
    return tok.decode(new).strip()