import llm import torch import torch.nn as nn from torch.nn import functional as F import math from dataclasses import dataclass import pickle import os @llm.hookimpl def register_models(register): register(CharGPT()) # @torch.jit.script # good to enable when not using torch.compile, disable when using (our default) def new_gelu(x): """ Implementation of the GELU activation function currently in Google BERT repo (identical to OpenAI GPT). Reference: Gaussian Error Linear Units (GELU) paper: https://arxiv.org/abs/1606.08415 """ return 0.5 * x * (1.0 + torch.tanh(math.sqrt(2.0 / math.pi) * (x + 0.044715 * torch.pow(x, 3.0)))) class LayerNorm(nn.Module): """ LayerNorm but with an optional bias. PyTorch doesn't support simply bias=False """ def __init__(self, ndim, bias): super().__init__() self.weight = nn.Parameter(torch.ones(ndim)) self.bias = nn.Parameter(torch.zeros(ndim)) if bias else None def forward(self, input): return F.layer_norm(input, self.weight.shape, self.weight, self.bias, 1e-5) class CausalSelfAttention(nn.Module): def __init__(self, config): super().__init__() assert config.n_embd % config.n_head == 0 # key, query, value projections for all heads, but in a batch self.c_attn = nn.Linear(config.n_embd, 3 * config.n_embd, bias=config.bias) # output projection self.c_proj = nn.Linear(config.n_embd, config.n_embd, bias=config.bias) # regularization self.attn_dropout = nn.Dropout(config.dropout) self.resid_dropout = nn.Dropout(config.dropout) self.n_head = config.n_head self.n_embd = config.n_embd self.dropout = config.dropout # flash attention make GPU go brrrrr but support is only in PyTorch nightly and still a bit scary self.flash = hasattr(torch.nn.functional, 'scaled_dot_product_attention') and self.dropout == 0.0 if not self.flash: # print("WARNING: using slow attention. Flash Attention atm needs PyTorch nightly and dropout=0.0") # causal mask to ensure that attention is only applied to the left in the input sequence self.register_buffer("bias", torch.tril(torch.ones(config.block_size, config.block_size)) .view(1, 1, config.block_size, config.block_size)) def forward(self, x): B, T, C = x.size() # batch size, sequence length, embedding dimensionality (n_embd) # calculate query, key, values for all heads in batch and move head forward to be the batch dim q, k ,v = self.c_attn(x).split(self.n_embd, dim=2) k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs) q = q.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs) v = v.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs) # causal self-attention; Self-attend: (B, nh, T, hs) x (B, nh, hs, T) -> (B, nh, T, T) if self.flash: # efficient attention using Flash Attention CUDA kernels y = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=self.dropout, is_causal=True) else: # manual implementation of attention att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(k.size(-1))) att = att.masked_fill(self.bias[:,:,:T,:T] == 0, float('-inf')) att = F.softmax(att, dim=-1) att = self.attn_dropout(att) y = att @ v # (B, nh, T, T) x (B, nh, T, hs) -> (B, nh, T, hs) y = y.transpose(1, 2).contiguous().view(B, T, C) # re-assemble all head outputs side by side # output projection y = self.resid_dropout(self.c_proj(y)) return y class MLP(nn.Module): def __init__(self, config): super().__init__() self.c_fc = nn.Linear(config.n_embd, 4 * config.n_embd, bias=config.bias) self.c_proj = nn.Linear(4 * config.n_embd, config.n_embd, bias=config.bias) self.dropout = nn.Dropout(config.dropout) def forward(self, x): x = self.c_fc(x) x = new_gelu(x) x = self.c_proj(x) x = self.dropout(x) return x class Block(nn.Module): def __init__(self, config): super().__init__() self.ln_1 = LayerNorm(config.n_embd, bias=config.bias) self.attn = CausalSelfAttention(config) self.ln_2 = LayerNorm(config.n_embd, bias=config.bias) self.mlp = MLP(config) def forward(self, x): x = x + self.attn(self.ln_1(x)) x = x + self.mlp(self.ln_2(x)) return x @dataclass class GPTConfig: block_size: int = 2048 vocab_size: int = 50304 # GPT-2 vocab_size of 50257, padded up to nearest multiple of 64 for efficiency n_layer: int = 12 n_head: int = 12 n_embd: int = 768 dropout: float = 0.0 bias: bool = True # True: bias in Linears and LayerNorms, like GPT-2. False: a bit better and faster class GPT(nn.Module): def __init__(self, config): super().__init__() assert config.vocab_size is not None assert config.block_size is not None self.config = config self.transformer = nn.ModuleDict(dict( wte = nn.Embedding(config.vocab_size, config.n_embd), wpe = nn.Embedding(config.block_size, config.n_embd), drop = nn.Dropout(config.dropout), h = nn.ModuleList([Block(config) for _ in range(config.n_layer)]), ln_f = LayerNorm(config.n_embd, bias=config.bias), )) self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False) # with weight tying when using torch.compile() some warnings get generated: # "UserWarning: functional_call was passed multiple values for tied weights. # This behavior is deprecated and will be an error in future versions" # not 100% sure what this is, so far seems to be harmless. TODO investigate self.transformer.wte.weight = self.lm_head.weight # https://paperswithcode.com/method/weight-tying # init all weights self.apply(self._init_weights) # apply special scaled init to the residual projections, per GPT-2 paper for pn, p in self.named_parameters(): if pn.endswith('c_proj.weight'): torch.nn.init.normal_(p, mean=0.0, std=0.02/math.sqrt(2 * config.n_layer)) # report number of parameters print("number of parameters: %.2fM" % (self.get_num_params()/1e6,)) def get_num_params(self, non_embedding=True): """ Return the number of parameters in the model. For non-embedding count (default), the position embeddings get subtracted. The token embeddings would too, except due to the parameter sharing these params are actually used as weights in the final layer, so we include them. """ n_params = sum(p.numel() for p in self.parameters()) if non_embedding: n_params -= self.transformer.wpe.weight.numel() return n_params def reset_parameters(self): # Initialize weights using Glorot initialization for param in self.parameters(): if param.dim() > 1: torch.nn.init.xavier_uniform_(param) def _init_weights(self, module): if isinstance(module, nn.Linear): torch.nn.init.normal_(module.weight, mean=0.0, std=0.02) if module.bias is not None: torch.nn.init.zeros_(module.bias) elif isinstance(module, nn.Embedding): torch.nn.init.normal_(module.weight, mean=0.0, std=0.02) def forward(self, idx, targets=None): device = idx.device b, t = idx.size() assert t <= self.config.block_size, f"Cannot forward sequence of length {t}, block size is only {self.config.block_size}" pos = torch.arange(0, t, dtype=torch.long, device=device).unsqueeze(0) # shape (1, t) # forward the GPT model itself tok_emb = self.transformer.wte(idx) # token embeddings of shape (b, t, n_embd) pos_emb = self.transformer.wpe(pos) # position embeddings of shape (1, t, n_embd) x = self.transformer.drop(tok_emb + pos_emb) for block in self.transformer.h: x = block(x) x = self.transformer.ln_f(x) if targets is not None: # if we are given some desired targets also calculate the loss logits = self.lm_head(x) loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1), ignore_index=-1) else: # inference-time mini-optimization: only forward the lm_head on the very last position logits = self.lm_head(x[:, [-1], :]) # note: using list [-1] to preserve the time dim loss = None return logits, loss @torch.no_grad() def generate_streaming(self, idx, max_new_tokens, temperature=1.0, top_k=None): """ Take a conditioning sequence of indices idx (LongTensor of shape (b,t)) and complete the sequence max_new_tokens times, feeding the predictions back into the model each time. Yield the generated indices one at a time rather than concatenating them into a single tensor. Most likely you'll want to make sure to be in model.eval() mode of operation for this. """ max_idx_length = self.config.block_size for _ in range(max_new_tokens): # if the sequence context is growing too long we must crop it at block_size idx_cond = idx if idx.size(1) <= max_idx_length else idx[:, -max_idx_length:] # forward the model to get the logits for the index in the sequence logits, _ = self(idx_cond) # pluck the logits at the final step and scale by desired temperature logits = logits[:, -1, :] / temperature # optionally crop the logits to only the top k options if top_k is not None: v, _ = torch.topk(logits, min(top_k, logits.size(-1))) logits[logits < v[:, [-1]]] = -float('Inf') # apply softmax to convert logits to (normalized) probabilities probs = F.softmax(logits, dim=-1) # sample from the distribution idx_next = torch.multinomial(probs, num_samples=1) # yield the next index # append sampled index to the running sequence and continue idx = torch.cat((idx, idx_next), dim=1) yield idx_next.item() def remove_caseifer(text): new_text = "" i = 0 while i < len(text): if text[i] == "↨": if i+1 < len(text): new_text += text[i+1].upper() i += 1 else: pass # skip this index else: new_text += text[i] i += 1 return new_text def add_caseifer(text): # Define your set of acceptable characters (original + keys from replace_map + replace_values) #chars = "\n\"\t' &@!$#,/\\+=-<>*%.…_:;[]}{()^?0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz§↨©®™¶¥¼°½¾«»£βθ♪ƒ~±¤º·\x8f€¢" tokenlist = "\n\t\x8f !#$%&()*+,-./:;<=>?@[\]^_{|}~§↨©®™¶¥¼°½¾«»£βθ♪ƒ±¤º·€¢\"'…0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz" upperlist = set("ABCDEFGHIJKLMNOPQRSTUVWXYZ") new_text = [] for char in text: if char in tokenlist: if char in upperlist: new_text.append("↨" + char.lower()) else: new_text.append(char) else: pass return "".join(new_text) model_dir = '16bit' device = 'cuda' dtype = 'bfloat16' class CharGPT(llm.Model): model_id = "chargpt" def execute(self, prompt, stream, response, conversation): torch.backends.cuda.matmul.allow_tf32 = True torch.backends.cudnn.allow_tf32 = True device_type = 'cuda' if 'cuda' in device else 'cpu' ptdtype = {'float32': torch.float32, 'bfloat16': torch.bfloat16, 'float16': torch.float16}[dtype] ctx = nullcontext() if device_type == 'cpu' else torch.amp.autocast(device_type=device_type, dtype=ptdtype) max_new_tokens = 2048 # number of tokens generated in each sample temperature = 0.8 # 1.0 = no change, < 1.0 = less random, > 1.0 = more random, in predictions top_k = 24 # retain only the top_k most likely tokens, clamp others to have 0 probability ckpt_path = os.path.join(model_dir, 'ckpt.pt') checkpoint = torch.load(ckpt_path, map_location=device) gptconf = GPTConfig(**checkpoint['model_args']) model = GPT(gptconf) state_dict = checkpoint['model'] unwanted_prefix = '_orig_mod.' for k,v in list(state_dict.items()): if k.startswith(unwanted_prefix): state_dict[k[len(unwanted_prefix):]] = state_dict.pop(k) model.load_state_dict(state_dict) model.eval() model.to(device) meta_path = os.path.join(model_dir, 'meta.pkl') with open(meta_path, 'rb') as f: meta = pickle.load(f) # TODO want to make this more general to arbitrary encoder/decoder schemes stoi, itos = meta['stoi'], meta['itos'] encode = lambda s: [stoi[c] for c in s] decode = lambda l: ''.join([itos[i] for i in l]) text = prompt.prompt shift = False # generated_text = '' start_ids = encode(add_caseifer(text)) x = (torch.tensor(start_ids, dtype=torch.long, device=device)[None, ...]) print(text, end='', flush=True) for idx_next in model.generate_streaming(x, max_new_tokens, temperature=temperature, top_k=top_k): # convert the index to a character and print it to the screen char = decode([idx_next]) # check for newline character if char == '§': # append the completed line to the list or print it to the screen # generated_sequences.append(generated_text) # reset the generated text for the next line # generated_data = generated_text # generated_text = '' break # append the character to the generated text if shift: # generated_text += char.upper() yield char.upper()# + '' # print(char.upper(), end='', flush=True) shift = False elif char == '↨': shift = True else: # generated_text += char yield char# + '' #print(char, end='', flush=True)