import torch import torch.nn as nn import torch.nn.functional as F import math from dataclasses import dataclass from typing import Optional, Tuple import tiktoken import os import numpy as np torch.set_float32_matmul_precision('high') def load_tokens(filename): npt = np.load(filename) npt = npt.astype(np.int32) # added after video ptt = torch.tensor(npt, dtype=torch.long) return ptt class DataLoaderLite: def __init__(self, B, T, split): self.B = B self.T = T assert split in {'train', 'val'} # get the shard filenames data_root = "tinystories" shards = os.listdir(data_root) shards = [s for s in shards if split in s] shards = sorted(shards) shards = [os.path.join(data_root, s) for s in shards] self.shards = shards assert len(shards) > 0, f"no shards found for split {split}" print(f"found {len(shards)} shards for split {split}") self.reset() def reset(self): # state, init at shard zero self.current_shard = 0 self.tokens = load_tokens(self.shards[self.current_shard]) self.current_position = self.B * self.T def next_batch(self): B, T = self.B, self.T buf = self.tokens[self.current_position : self.current_position+B*T+1] x = (buf[:-1]).view(B, T) # inputs y = (buf[1:]).view(B, T) # targets # advance the position in the tensor self.current_position += B * T # if loading the next batch would be out of bounds, advance to next shard if self.current_position + (B * T ) > len(self.tokens): self.current_shard = (self.current_shard + 1) % len(self.shards) self.tokens = load_tokens(self.shards[self.current_shard]) self.current_position = B * T return x, y @dataclass # the hyperparameters class ModelArgs: dim: int = 768 n_layers: int = 16 n_heads: int = 16 n_kv_heads: Optional[int] = 4 vocab_size: int = 50304 multiple_of: int = 256 ffn_dim_multiplier: Optional[float] = None norm_eps: float = 1e-5 rope_theta: float = 50000 max_batch_size: int = 4 max_seq_len: int = 1024 device: str = 'cuda' if torch.cuda.is_available() else 'cpu' dropout_rate: float = 0.1 params = ModelArgs() class RMSNorm(torch.nn.Module): def __init__(self, dim: int, eps: float = 1e-6): super().__init__() self.eps = eps self.weight = nn.Parameter(torch.ones(dim)) def _norm(self, x): return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps) def forward(self, x): output = self._norm(x.float()).type_as(x) return output * self.weight def precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0): freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim)) t = torch.arange(end, device=freqs.device, dtype=torch.float32) freqs = torch.outer(t, freqs) freqs_cis = torch.polar(torch.ones_like(freqs), freqs) # complex64 return freqs_cis.to(params.device) def reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor): ndim = x.ndim assert 0 <= 1 < ndim assert freqs_cis.shape == (x.shape[1], x.shape[-1]), f'freqs_cis.shape {freqs_cis.shape} != (x.shape[1], x.shape[-1]) {(x.shape[1], x.shape[-1])}' shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)] return freqs_cis.view(*shape) def apply_rotary_emb( xq: torch.Tensor, xk: torch.Tensor, freqs_cis: torch.Tensor, ) -> Tuple[torch.Tensor, torch.Tensor]: xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2)) xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2)) freqs_cis = reshape_for_broadcast(freqs_cis, xq_) xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3) xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3) return xq_out.type_as(xq), xk_out.type_as(xk) def repeat_kv(x: torch.Tensor, n_rep: int) -> torch.Tensor: """torch.repeat_interleave(x, dim=2, repeats=n_rep)""" bs, seqlen, n_kv_heads, head_dim = x.shape if n_rep == 1: return x return ( x[:, :, :, None, :] .expand(bs, seqlen, n_kv_heads, n_rep, head_dim) .reshape(bs, seqlen, n_kv_heads * n_rep, head_dim) ) class Attention(nn.Module): def __init__(self, args: ModelArgs): super().__init__() self.n_heads = args.n_heads self.n_kv_heads = args.n_heads if args.n_kv_heads is None else args.n_kv_heads self.n_rep = args.n_heads // self.n_kv_heads self.head_dim = args.dim // args.n_heads self.wq = nn.Linear(args.dim, args.n_heads * self.head_dim, bias=False) self.wk = nn.Linear(args.dim, self.n_kv_heads * self.head_dim, bias=False) self.wv = nn.Linear(args.dim, self.n_kv_heads * self.head_dim, bias=False) self.wo = nn.Linear(args.n_heads * self.head_dim, args.dim, bias=False) self.cache_k = torch.zeros( (args.max_batch_size, args.max_seq_len, self.n_kv_heads, self.head_dim), requires_grad = False ).to(args.device) self.cache_v = torch.zeros( (args.max_batch_size, args.max_seq_len, self.n_kv_heads, self.head_dim), requires_grad = False ).to(args.device) def forward( self, x: torch.Tensor, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], start_pos: int = None, ): bsz, seqlen, _ = x.shape xq, xk, xv = self.wq(x), self.wk(x), self.wv(x) xq = xq.view(bsz, seqlen, self.n_heads, self.head_dim) xk = xk.view(bsz, seqlen, self.n_kv_heads, self.head_dim) xv = xv.view(bsz, seqlen, self.n_kv_heads, self.head_dim) xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis) if start_pos is not None: # if we're performing inference, use kv caching self.cache_k = self.cache_k.to(xq) self.cache_v = self.cache_v.to(xq) # set the values in our cache according to the current input self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv # grab our key and value matrixes which have a longer sequence length than our queries keys = self.cache_k[:bsz, : start_pos + seqlen] values = self.cache_v[:bsz, : start_pos + seqlen] else: # if we're training, do full sequence length keys, values = xk, xv keys = repeat_kv(keys, self.n_rep) values = repeat_kv(values, self.n_rep) xq = xq.transpose(1, 2) keys = keys.transpose(1, 2) values = values.transpose(1, 2) scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim) if mask is not None: scores = scores + mask scores = F.softmax(scores.float(), dim=-1).type_as(xq) output = torch.matmul(scores, values) output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1) return self.wo(output) class FeedForward(nn.Module): def __init__( self, dim: int, hidden_dim: int, multiple_of: int, ffn_dim_multiplier: Optional[float], ): super().__init__() # custom dim factor multiplier that ensures we're using a multiple of "multiple_of" for hardware efficiency reasons hidden_dim = int(2 * hidden_dim / 3) if ffn_dim_multiplier is not None: hidden_dim = int(ffn_dim_multiplier * hidden_dim) hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of) self.w1 = nn.Linear(dim, hidden_dim, bias=False) self.w2 = nn.Linear(hidden_dim, dim, bias=False) self.w3 = nn.Linear(dim, hidden_dim, bias=False) def forward(self, x): return self.w2(F.silu(self.w1(x)) * self.w3(x)) class TransformerBlock(nn.Module): def __init__(self, args: ModelArgs): super().__init__() self.n_heads = args.n_heads self.dim = args.dim self.head_dim = args.dim // args.n_heads self.attention = Attention(args) self.feed_forward = FeedForward( dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, ffn_dim_multiplier=args.ffn_dim_multiplier, ) self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps) self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps) self.dropout_rate = args.dropout_rate def forward( self, x: torch.Tensor, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], start_pos: int = None, training = False, ): h = x + F.dropout(self.attention(self.attention_norm(x), freqs_cis, mask, start_pos), p=self.dropout_rate, training=training) out = h + F.dropout(self.feed_forward(self.ffn_norm(h)), p=self.dropout_rate, training=training) return out class Moose(nn.Module): def __init__(self, params: ModelArgs): super().__init__() self.params = params self.vocab_size = params.vocab_size self.n_layers = params.n_layers self.max_seq_len = params.max_seq_len self.tok_embeddings = nn.Embedding(params.vocab_size, params.dim) self.layers = torch.nn.ModuleList() for _ in range(params.n_layers): self.layers.append(TransformerBlock(params)) self.norm = RMSNorm(params.dim, eps=params.norm_eps) self.output = nn.Linear( params.dim, params.vocab_size, bias=False) self.freqs_cis = precompute_freqs_cis( params.dim // params.n_heads, params.max_seq_len * 2, params.rope_theta,) # precompute the causal attention mask mask = torch.full((params.max_seq_len, params.max_seq_len), float("-inf"), device=params.device) mask = torch.triu(mask, diagonal=1) self.register_buffer('mask', mask) self.criterion = nn.CrossEntropyLoss() def forward(self, tokens: torch.Tensor, targets: torch.Tensor): bsz, seqlen = tokens.shape assert tokens.shape == targets.shape assert seqlen == self.max_seq_len h = self.tok_embeddings(tokens) freqs_cis = self.freqs_cis.to(h.device) freqs_cis = self.freqs_cis[:seqlen] for layer in self.layers: h = layer( h, freqs_cis, self.mask, start_pos = None, training = True ) h = self.norm(h) logits = self.output(h).float() loss = self.criterion( logits.view(bsz * seqlen, self.vocab_size), targets.reshape(bsz * seqlen)) return logits, loss @torch.inference_mode() def forward_inference(self, tokens: torch.Tensor, start_pos: int, max_context_window: int, ): _bsz, seqlen = tokens.shape h = self.tok_embeddings(tokens) self.freqs_cis = self.freqs_cis.to(h.device) freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen] mask = self.mask[:seqlen, :seqlen] mask = torch.hstack( [torch.zeros((seqlen, start_pos), device=tokens.device), mask] ).type_as(h) for layer in self.layers: h = layer( h, freqs_cis, mask, start_pos = start_pos ) h = self.norm(h) logits = self.output(h).float() return logits @torch.inference_mode() def Sampler( self, logits: torch.Tensor, temperature: float, top_p: float ) -> torch.Tensor: logits = logits[:,-1,:] logits.div_(temperature) probs = torch.softmax(logits, dim=-1, dtype=torch.float) probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True) probs_sum = torch.cumsum(probs_sort, dim=-1) top_ps_mask = (probs_sum - probs_sort) > top_p probs_sort = torch.where(top_ps_mask, 0, probs_sort) probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True)) probs = torch.gather(probs_sort, dim=-1, index=torch.argsort(probs_idx, dim=-1)) next_token_id = torch.multinomial(probs, num_samples=1) return next_token_id @torch.inference_mode() def generate( self, prompt: str, max_gen_len: int = 100, memory_saver_div: int = 1, temperature: float = 0.6, top_p: float = 0.9, ) -> str: assert ((memory_saver_div & (memory_saver_div-1)) == 0) & (memory_saver_div > 0), f'memory_saver_div {memory_saver_div} must be power of 2' max_context_window = self.max_seq_len // memory_saver_div enc = tiktoken.get_encoding('gpt2') tokens = enc.encode(prompt) tokens = torch.tensor(tokens, device=self.params.device) tokens = tokens.unsqueeze(0) if len(tokens.shape)==1 else tokens start_pos = max(tokens.shape[1] - max_context_window, 0) eot = enc._special_tokens['<|endoftext|>'] # end of text token while True: logits = self.forward_inference( tokens[:,-max_context_window:], start_pos = start_pos, max_context_window = max_context_window ) next_token = self.Sampler( logits = logits, temperature = temperature, top_p = top_p, ) tokens = torch.cat((tokens, next_token), dim=1) if next_token.item() == eot: break if tokens.shape[1] >= max_context_window: start_pos += 1 output = enc.decode(tokens.squeeze(0).tolist()) return output[:-13]