model.py

"""
Full definition of a GPT Language Model, all of it in this single file.
References:
1) the official GPT-2 TensorFlow implementation released by OpenAI:
https://github.com/openai/gpt-2/blob/master/src/model.py
2) huggingface/transformers PyTorch implementation:
https://github.com/huggingface/transformers/blob/main/src/transformers/models/gpt2/modeling_gpt2.py
"""

from datetime import datetime
import math
import inspect
import os
import uuid

import pandas as pd
from pydantic import BaseModel, ConfigDict
import torch
import torch.nn as nn
from torch.nn import functional as F
from transformers import PreTrainedTokenizerFast
from typing import Callable


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 >= 2.0
        self.flash = hasattr(torch.nn.functional, "scaled_dot_product_attention")
        if not self.flash:
            print(
                "WARNING: using slow attention. Flash Attention requires PyTorch >= 2.0"
            )
            # 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 if self.training else 0,
                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.gelu = nn.GELU()
        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 = self.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


class GPTConfig(BaseModel):
    block_size: int = 1024
    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
    tokenizer_file: str = 'resources/tokenizer.json'

    model_config = ConfigDict(extra='ignore')


class GPT(nn.Module):
    def __init__(self, config: GPTConfig):
        super().__init__()
        assert config.vocab_size is not None
        assert config.block_size is not None
        self.config = config
        self.tokenizer = PreTrainedTokenizerFast(tokenizer_file=config.tokenizer_file)
        self.end_token = self.tokenizer('[END]')['input_ids'][0]
        self.comma_token = self.tokenizer(',')['input_ids'][0]

        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 _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):
        # with torch.autograd.detect_anomaly():
        #     if torch.isnan(idx).any():
        #         print(f'NAN found!: {idx}')

        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)  # shape (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 (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

    def crop_block_size(self, block_size):
        # model surgery to decrease the block size if necessary
        # e.g. we may load the GPT2 pretrained model checkpoint (block size 1024)
        # but want to use a smaller block size for some smaller, simpler model
        assert block_size <= self.config.block_size
        self.config.block_size = block_size
        self.transformer.wpe.weight = nn.Parameter(
            self.transformer.wpe.weight[:block_size]
        )
        for block in self.transformer.h:
            if hasattr(block.attn, "bias"):
                block.attn.bias = block.attn.bias[:, :, :block_size, :block_size]

    @classmethod
    def from_pretrained(cls, model_type, override_args=None):
        assert model_type in {"gpt2", "gpt2-medium", "gpt2-large", "gpt2-xl"}
        override_args = override_args or {}  # default to empty dict
        # only dropout can be overridden see more notes below
        assert all(k == "dropout" for k in override_args)
        from transformers import GPT2LMHeadModel

        print("loading weights from pretrained gpt: %s" % model_type)

        # n_layer, n_head and n_embd are determined from model_type
        config_args = {
            "gpt2": dict(n_layer=12, n_head=12, n_embd=768),  # 124M params
            "gpt2-medium": dict(n_layer=24, n_head=16, n_embd=1024),  # 350M params
            "gpt2-large": dict(n_layer=36, n_head=20, n_embd=1280),  # 774M params
            "gpt2-xl": dict(n_layer=48, n_head=25, n_embd=1600),  # 1558M params
        }[model_type]
        print("forcing vocab_size=50257, block_size=1024, bias=True")
        config_args["vocab_size"] = 50257  # always 50257 for GPT model checkpoints
        config_args["block_size"] = 1024  # always 1024 for GPT model checkpoints
        config_args["bias"] = True  # always True for GPT model checkpoints
        # we can override the dropout rate, if desired
        if "dropout" in override_args:
            print(f"overriding dropout rate to {override_args['dropout']}")
            config_args["dropout"] = override_args["dropout"]
        # create a from-scratch initialized minGPT model
        config = GPTConfig(**config_args)
        model = GPT(config)
        sd = model.state_dict()
        sd_keys = sd.keys()
        sd_keys = [
            k for k in sd_keys if not k.endswith(".attn.bias")
        ]  # discard this mask / buffer, not a param

        # init a huggingface/transformers model
        model_hf = GPT2LMHeadModel.from_pretrained(model_type)
        sd_hf = model_hf.state_dict()

        # copy while ensuring all of the parameters are aligned and match in names and shapes
        sd_keys_hf = sd_hf.keys()
        sd_keys_hf = [
            k for k in sd_keys_hf if not k.endswith(".attn.masked_bias")
        ]  # ignore these, just a buffer
        sd_keys_hf = [
            k for k in sd_keys_hf if not k.endswith(".attn.bias")
        ]  # same, just the mask (buffer)
        transposed = [
            "attn.c_attn.weight",
            "attn.c_proj.weight",
            "mlp.c_fc.weight",
            "mlp.c_proj.weight",
        ]
        # basically the openai checkpoints use a "Conv1D" module, but we only want to use a vanilla Linear
        # this means that we have to transpose these weights when we import them
        assert len(sd_keys_hf) == len(
            sd_keys
        ), f"mismatched keys: {len(sd_keys_hf)} != {len(sd_keys)}"
        for k in sd_keys_hf:
            if any(k.endswith(w) for w in transposed):
                # special treatment for the Conv1D weights we need to transpose
                assert sd_hf[k].shape[::-1] == sd[k].shape
                with torch.no_grad():
                    sd[k].copy_(sd_hf[k].t())
            else:
                # vanilla copy over the other parameters
                assert sd_hf[k].shape == sd[k].shape
                with torch.no_grad():
                    sd[k].copy_(sd_hf[k])

        return model

    def configure_optimizers(self, weight_decay, learning_rate, betas, device_type):
        # start with all of the candidate parameters
        param_dict = {pn: p for pn, p in self.named_parameters()}
        # filter out those that do not require grad
        param_dict = {pn: p for pn, p in param_dict.items() if p.requires_grad}
        # create optim groups. Any parameters that is 2D will be weight decayed, otherwise no.
        # i.e. all weight tensors in matmuls + embeddings decay, all biases and layernorms don't.
        decay_params = [p for n, p in param_dict.items() if p.dim() >= 2]
        nodecay_params = [p for n, p in param_dict.items() if p.dim() < 2]
        optim_groups = [
            {"params": decay_params, "weight_decay": weight_decay},
            {"params": nodecay_params, "weight_decay": 0.0},
        ]
        num_decay_params = sum(p.numel() for p in decay_params)
        num_nodecay_params = sum(p.numel() for p in nodecay_params)
        print(
            f"num decayed parameter tensors: {len(decay_params)}, with {num_decay_params:,} parameters"
        )
        print(
            f"num non-decayed parameter tensors: {len(nodecay_params)}, with {num_nodecay_params:,} parameters"
        )
        # Create AdamW optimizer and use the fused version if it is available
        fused_available = "fused" in inspect.signature(torch.optim.AdamW).parameters
        use_fused = fused_available and device_type == "cuda"
        extra_args = dict(fused=True) if use_fused else dict()
        optimizer = torch.optim.AdamW(
            optim_groups, lr=learning_rate, betas=betas, **extra_args
        )
        print(f"using fused AdamW: {use_fused}")

        return optimizer

    def estimate_mfu(self, fwdbwd_per_iter, dt):
        """estimate model flops utilization (MFU) in units of A100 bfloat16 peak FLOPS"""
        # first estimate the number of flops we do per iteration.
        # see PaLM paper Appendix B as ref: https://arxiv.org/abs/2204.02311
        N = self.get_num_params()
        cfg = self.config
        L, H, Q, T = cfg.n_layer, cfg.n_head, cfg.n_embd // cfg.n_head, cfg.block_size
        flops_per_token = 6 * N + 12 * L * H * Q * T
        flops_per_fwdbwd = flops_per_token * T
        flops_per_iter = flops_per_fwdbwd * fwdbwd_per_iter
        # express our flops throughput as ratio of A100 bfloat16 peak flops
        flops_achieved = flops_per_iter * (1.0 / dt)  # per second
        flops_promised = 312e12  # A100 GPU bfloat16 peak flops is 312 TFLOPS
        mfu = flops_achieved / flops_promised
        return mfu

    @property
    def device(self) -> str:
        # assign model inputs to the right device
        return next(self.lm_head.parameters()).device.type

    @torch.no_grad()
    def generate(
            self,
            idx: torch.Tensor,
            max_new_tokens: int = 12,
            temperature: float = 0.0,
            topn: int = 100,
            pruning_ratio: float = 4,
            pruning_offset: float = 5,
            log_file: str | None = None,
            on_iteration: Callable = None,
    ) -> torch.Tensor:

        if topn <= 0:
            raise ValueError('topn should be greater than 0')

        if not 0 < max_new_tokens <= 20:
            raise ValueError('max_new_tokens should be in (0, 20]')

        run_uuid = uuid.uuid4()

        idx = idx.to(self.device)
        sequences = idx.unsqueeze(0)

        probabilities = torch.tensor([1.], device=self.device)

        finished_sequences = torch.tensor([], device=self.device)
        finished_probs = torch.tensor([], device=self.device)

        # compute number of sequences to pass to each iteration
        sequences_per_iter = round(pruning_offset + topn / pruning_ratio)

        for i in range(max_new_tokens):
            if on_iteration is not None:
                on_iteration()

            # trim the sequences down to block size
            sequences = sequences[:, -self.config.block_size:]

            # inference the model
            logits, _ = self(sequences)
            logits = logits.squeeze(1)

            # take N most probable next tokens for each sequence
            output_probs = F.softmax(logits, dim=-1)
            new_sequence_probs = output_probs * probabilities.unsqueeze(1)

            # remove finished sequences (after end token) and cache their probs
            if i > 0:
                # feature to add: we should not add subdomain in input to the finished sequences
                comma_token_probs = new_sequence_probs[:, self.comma_token]
                end_token_probs = new_sequence_probs[:, self.end_token]
                _finish_probs = end_token_probs + comma_token_probs

                finished_sequences = torch.cat((finished_sequences, sequences))
                finished_probs = torch.cat((finished_probs, _finish_probs), dim=-1)

            # remove sequences and tokens with a probability that is too low
            if len(finished_sequences) > topn:
                # torch.kthvalue is not implemented on MPS, so we use topk
                lowest_viable_probability = torch.topk(finished_probs, topn).values[-1]
                viable_sequences = probabilities > lowest_viable_probability

                if viable_sequences.sum() == 0:
                    break

                # remove sequences with a too low probability
                sequences = sequences[viable_sequences]
                probabilities = probabilities[viable_sequences]
                logits = logits[viable_sequences]
                new_sequence_probs = new_sequence_probs[viable_sequences]

                # remove tokens that would generate sequences with too low probability
                token_mask = new_sequence_probs < lowest_viable_probability
                if token_mask.sum() == 0:
                    break

                new_sequence_probs[token_mask] = 0
                logits[token_mask] = 0

            # do not sample the end token or comma token for the next iter
            new_sequence_probs[:, self.end_token] = 0
            new_sequence_probs[:, self.comma_token] = 0

            # number of sequences to pass to next iteration
            num_nonzero_probs = torch.count_nonzero(new_sequence_probs).item()
            num_seqs_next_iter = min(sequences_per_iter, num_nonzero_probs)

            if num_seqs_next_iter == 0:
                break

            if temperature == 0:  # select most likely tokens for next iteration
                new_sequence_probs = new_sequence_probs.flatten()
                _, idx_next = torch.topk(new_sequence_probs, num_seqs_next_iter)

            else:  # sample tokens for next iteration
                # recalculate probabilities using temperature
                scaled_logits = logits / (temperature+1e-1)
                probs_with_temp = F.softmax(scaled_logits, dim=-1)
                probs_with_temp = probs_with_temp * probabilities.unsqueeze(1)

                probs_with_temp[:, self.end_token] = 0
                probs_with_temp[:, self.comma_token] = 0

                # sample tokens for next iteration
                probs_with_temp = probs_with_temp.flatten()
                probs_with_temp[probs_with_temp < 0] = 0
                idx_next = torch.multinomial(probs_with_temp, num_seqs_next_iter)

            # add the sampled tokens to the end of each sequence
            sequence_idx = idx_next // self.config.vocab_size
            token_values = idx_next % self.config.vocab_size

            sequences = sequences[sequence_idx]
            sequences = torch.cat([sequences, token_values.unsqueeze(1)], dim=-1)
            probabilities = new_sequence_probs.flatten()[idx_next]

            if log_file is not None:
                _, current_best_idx = torch.topk(finished_probs, min(topn, len(finished_probs)))
                current_best = finished_sequences[current_best_idx]
                self.log_generation_data(
                    log_file=log_file,
                    run_id=run_uuid,
                    topn=topn,
                    x=idx,
                    iteration=i,
                    probabilities=probabilities,
                    current_preds=current_best,
                    finished_probs=finished_probs,
                )

        # take the highest scoring sequences for the next iteration
        _, final_indices = torch.topk(finished_probs, topn)
        final_sequences = finished_sequences[final_indices]

        return final_sequences

    def log_generation_data(
            self,
            log_file: str,
            run_id: uuid.UUID,
            iteration: int,
            topn: int,
            x: torch.Tensor,
            probabilities: torch.Tensor,
            current_preds: torch.Tensor,
            finished_probs: torch.Tensor,
    ):
        # use this in every iteration of the generate method to collect data for analysis

        # # turn into list of ints
        # current_preds = current_preds.int().tolist()
        #
        # # turn into list of strings
        # current_preds = [
        #     self.tokenizer.decode(pred)
        #     .replace(" ", "")
        #     .rsplit("[DELIM]", 1)[1]
        #     for pred in current_preds
        # ]
        #
        # # turn into comma separated strings
        # current_preds = ','.join(current_preds)
        #
        # x = x.int().tolist()
        # x = self.tokenizer.decode(x).replace('[PAD]', '').replace(' ', '')

        if len(finished_probs) > topn:
            topnth_finished_prob = torch.topk(finished_probs, topn).values[-1].item()
        else:
            topnth_finished_prob = 0

        largest_prob = probabilities.max().item()

        new_row = [{
            'time': datetime.now().strftime('%Y-%m-%d %H:%M:%S.%f'),
            'run_id': str(run_id),
            'topn': topn,
            'iteration': iteration,
            'largest_prob': largest_prob,
            'topnth_finished_prob': topnth_finished_prob,
            # 'x': x,
            # 'probabilities': probabilities.sum().item(),
            # 'finished_probabilities': finished_probs.sum().item(),
            # 'finished_sequences': current_preds,
        }]
        df_new_row = pd.DataFrame(new_row)

        if os.path.exists(log_file):
            df = pd.read_csv(log_file, index_col=0)
            df = pd.concat([df, df_new_row], ignore_index=True)
        else:
            df = df_new_row

        df.to_csv(log_file)

    def save_checkpoint(
        self, path, optimizer=None, iter_num=None, best_val_loss=None, config=None
    ):
        optimizer = {} if not optimizer else optimizer.state_dict()
        iter_num = {} if not iter_num else {"iter_num": iter_num}
        best_val_loss = {} if not best_val_loss else {"best_val_loss": best_val_loss}
        config = {} if not config else {"config": config}
        checkpoint = {
            "model": self.state_dict(),
            "model_args": dict(self.config),
            **optimizer,
            **iter_num,
            **best_val_loss,
            **config,
        }
        torch.save(checkpoint, path)

    @staticmethod
    def from_checkpoint(
        path: str,
        return_train_params: bool = False,
        device: str = 'cpu',
        tokenizer_path: str | None = None,
    ):
        checkpoint = torch.load(path, map_location=device, weights_only=True)

        config = GPTConfig(**checkpoint["model_args"])
        if tokenizer_path:
            config.tokenizer_file = tokenizer_path
        model = GPT(config)
        state_dict = checkpoint["model"]

        # fix the keys of the state dictionary :(
        # honestly no idea how checkpoints sometimes get this prefix, have to debug more
        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.to(device)

        if not return_train_params:
            return model

        iter_num = checkpoint["iter_num"]
        best_val_loss = checkpoint["best_val_loss"]
        optim_state = checkpoint["optimizer"]

        assert isinstance(iter_num, int)
        assert isinstance(best_val_loss, torch.Tensor)
        assert isinstance(optim_state, dict)

        return model, iter_num, best_val_loss, optim_state