generate.py

"""
LUNA 100M — Text Generation / Interactive Chat
Usage:
    python generate.py                                          # interactive REPL
    python generate.py --prompt "The future of AI is"          # single prompt
    python generate.py --ckpt Base/out/luna_100m/latest.pt     # custom checkpoint
    python generate.py --max_new 200 --temp 0.8 --top_p 0.9   # tune generation
"""

import sys
import math
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
from pathlib import Path


# ─── Model (must match train.py exactly) ──────────────────────────────────────

class RotaryEmbedding(nn.Module):
    def __init__(self, dim, max_seq_len=1024):
        super().__init__()
        inv_freq = 1.0 / (10000 ** (torch.arange(0, dim, 2).float() / dim))
        self.register_buffer("inv_freq", inv_freq)
        t = torch.arange(max_seq_len).float()
        freqs = torch.einsum("i,j->ij", t, inv_freq)
        emb = torch.cat([freqs, freqs], dim=-1)
        self.register_buffer("cos_cached", emb.cos())
        self.register_buffer("sin_cached", emb.sin())

    def forward(self, seq_len):
        return self.cos_cached[:seq_len], self.sin_cached[:seq_len]


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


def apply_rotary(x, cos, sin):
    c = cos.unsqueeze(0).unsqueeze(0)
    s = sin.unsqueeze(0).unsqueeze(0)
    return x * c + rotate_half(x) * s


class CausalSelfAttention(nn.Module):
    def __init__(self, n_embd, n_head, block_size, rotary_pct=0.25):
        super().__init__()
        self.n_head = n_head
        self.head_dim = n_embd // n_head
        self.rot_dim = int(self.head_dim * rotary_pct)
        self.c_attn = nn.Linear(n_embd, 3 * n_embd, bias=True)
        self.c_proj = nn.Linear(n_embd, n_embd, bias=True)
        self.rotary = RotaryEmbedding(self.rot_dim, block_size)

    def forward(self, x):
        B, T, C = x.size()
        qkv = self.c_attn(x).reshape(B, T, 3, self.n_head, self.head_dim).permute(2, 0, 3, 1, 4)
        q, k, v = qkv.unbind(0)
        cos, sin = self.rotary(T)
        q = torch.cat([apply_rotary(q[..., :self.rot_dim], cos, sin), q[..., self.rot_dim:]], dim=-1)
        k = torch.cat([apply_rotary(k[..., :self.rot_dim], cos, sin), k[..., self.rot_dim:]], dim=-1)
        y = F.scaled_dot_product_attention(q, k, v, is_causal=True)
        return self.c_proj(y.transpose(1, 2).contiguous().view(B, T, C))


class MLP(nn.Module):
    def __init__(self, n_embd):
        super().__init__()
        self.fc   = nn.Linear(n_embd, 4 * n_embd, bias=True)
        self.gelu = nn.GELU()
        self.proj = nn.Linear(4 * n_embd, n_embd, bias=True)

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


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

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


class LUNAModel(nn.Module):
    def __init__(self, vocab_size=50304, block_size=1024,
                 n_layer=10, n_embd=768, n_head=12):
        super().__init__()
        self.block_size = block_size
        self.wte    = nn.Embedding(vocab_size, n_embd)
        self.blocks = nn.ModuleList([Block(n_embd, n_head, block_size) for _ in range(n_layer)])
        self.ln_f   = nn.LayerNorm(n_embd)
        self.lm_head = nn.Linear(n_embd, vocab_size, bias=False)
        self.lm_head.weight = self.wte.weight  # tied

    def forward(self, idx):
        x = self.wte(idx)
        for block in self.blocks:
            x = block(x)
        return self.lm_head(self.ln_f(x))


# ─── Generation ───────────────────────────────────────────────────────────────

@torch.no_grad()
def generate(model, input_ids, max_new=200, temperature=0.8,
             top_p=0.9, top_k=50, repetition_penalty=1.1, device="cpu"):
    model.eval()
    ids = input_ids.clone().to(device)
    generated = []

    for _ in range(max_new):
        # Crop to block_size
        ctx = ids[:, -model.block_size:]
        logits = model(ctx)        # (1, T, V)
        logits = logits[:, -1, :]  # last token

        # Repetition penalty
        if repetition_penalty != 1.0:
            for token_id in set(ids[0].tolist()):
                logits[0, token_id] /= repetition_penalty

        logits = logits / max(temperature, 1e-8)

        # Top-k
        if top_k > 0:
            vals, _ = torch.topk(logits, min(top_k, logits.size(-1)))
            logits[logits < vals[:, -1:]] = -float("inf")

        # Top-p (nucleus)
        probs = torch.softmax(logits, dim=-1)
        if top_p < 1.0:
            sorted_probs, sorted_idx = torch.sort(probs, descending=True)
            cum = torch.cumsum(sorted_probs, dim=-1)
            mask = cum - sorted_probs > top_p
            sorted_probs[mask] = 0.0
            sorted_probs /= sorted_probs.sum()
            next_token = sorted_idx[0, torch.multinomial(sorted_probs[0], 1)]
        else:
            next_token = torch.multinomial(probs[0], 1)

        ids = torch.cat([ids, next_token.view(1, 1)], dim=1)
        generated.append(next_token.item())

        # Stop at EOS
        if next_token.item() == 50276:
            break

    return generated


# ─── Load ─────────────────────────────────────────────────────────────────────

def load_model(ckpt_path: str, device: str):
    print(f"Loading checkpoint: {ckpt_path}")
    ckpt = torch.load(ckpt_path, map_location="cpu", weights_only=True)

    # Handle both raw state_dict and {'model': ...} wrappers
    state = ckpt["model"] if "model" in ckpt else ckpt
    step = ckpt.get("step", "?")
    tokens = ckpt.get("tokens_seen", 0)
    print(f"  Step: {step}  |  Tokens seen: {tokens:,}")

    model = LUNAModel()
    model.load_state_dict(state, strict=True)
    model = model.to(device)
    model.eval()
    print(f"  Parameters: {sum(p.numel() for p in model.parameters()):,}")
    return model


def load_tokenizer(tok_dir: str):
    try:
        from transformers import AutoTokenizer
        tok = AutoTokenizer.from_pretrained(tok_dir)
        print(f"  Tokenizer: {tok_dir}  (vocab {tok.vocab_size})")
        return tok
    except Exception as e:
        print(f"  ERROR loading tokenizer: {e}")
        print("  Install: pip install transformers")
        sys.exit(1)


# ─── Entry ────────────────────────────────────────────────────────────────────

def parse_args():
    p = argparse.ArgumentParser(description="LUNA 100M - Text Generation")
    p.add_argument("--ckpt",    default="Base/out/luna_100m/latest.pt")
    p.add_argument("--tok_dir", default="Base/checkpoints/EleutherAI/pythia-160m")
    p.add_argument("--prompt",  default=None,  help="Single prompt (else interactive)")
    p.add_argument("--max_new", type=int,   default=200)
    p.add_argument("--temp",    type=float, default=0.8)
    p.add_argument("--top_p",   type=float, default=0.9)
    p.add_argument("--top_k",   type=int,   default=50)
    p.add_argument("--rep_pen", type=float, default=1.1, help="Repetition penalty")
    p.add_argument("--device",  default="auto")
    return p.parse_args()


def run_prompt(model, tokenizer, prompt, args, device):
    ids = tokenizer.encode(prompt, return_tensors="pt")
    print(f"\n{'='*60}")
    print(f"PROMPT: {prompt}")
    print(f"{'='*60}")
    print(prompt, end="", flush=True)

    new_ids = generate(
        model, ids,
        max_new=args.max_new,
        temperature=args.temp,
        top_p=args.top_p,
        top_k=args.top_k,
        repetition_penalty=args.rep_pen,
        device=device,
    )
    output = tokenizer.decode(new_ids, skip_special_tokens=True)
    print(output)
    print(f"{'='*60}")
    print(f"Generated {len(new_ids)} tokens")


def main():
    args = parse_args()

    if args.device == "auto":
        device = "cuda" if torch.cuda.is_available() else "cpu"
    else:
        device = args.device
    print(f"\nDevice: {device}")

    model    = load_model(args.ckpt, device)
    tokenizer = load_tokenizer(args.tok_dir)

    if args.prompt:
        run_prompt(model, tokenizer, args.prompt, args, device)
        return

    # Interactive REPL
    print(f"\n{'='*60}")
    print("  LUNA 100M - Interactive Generation")
    print(f"  Checkpoint: {args.ckpt}")
    print(f"  max_new={args.max_new}  temp={args.temp}  top_p={args.top_p}  top_k={args.top_k}")
    print("  Type your prompt and press Enter. Ctrl+C to exit.")
    print(f"{'='*60}\n")

    while True:
        try:
            prompt = input(">>> ").strip()
            if not prompt:
                continue
            run_prompt(model, tokenizer, prompt, args, device)
        except KeyboardInterrupt:
            print("\nBye!")
            break
        except EOFError:
            break


if __name__ == "__main__":
    main()