Upload GPT

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,18 @@
+{
+  "architectures": [
+    "GPT"
+  ],
+  "auto_map": {
+    "AutoConfig": "modeling_gpt.GPTConfig",
+    "AutoModelForCausalLM": "modeling_gpt.GPT"
+  },
+  "dtype": "float32",
+  "emb_dim": 768,
+  "head_num": 12,
+  "layer_num": 12,
+  "max_context_len": 1024,
+  "micro_batch": 1,
+  "model_type": "custom_gpt2",
+  "transformers_version": "5.12.0",
+  "vocab_size": 50304
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "transformers_version": "5.12.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84b73804e3dda3fb92ef6f2bc7d214857b0f6ede6542b2dc4cb682287633cd68
+size 548251520

modeling_gpt.py

@@ -0,0 +1,249 @@
+import inspect
+
+import torch
+from torch import nn as nn, optim
+from torch.nn import functional as F
+import torch
+from torch.nn.attention import sdpa_kernel, SDPBackend
+from transformers import (
+    GenerationMixin,
+    PreTrainedConfig,
+    PreTrainedModel,
+)
+from transformers.modeling_outputs import CausalLMOutput
+
+class GPTConfig(PreTrainedConfig):
+    """
+    Attributes
+    ----------
+    vocab_size : int
+        語彙数
+    max_context_len : int, default = 1024
+        最大コンテキスト長
+    layer_num : int
+        Transformerブロックの層数
+    head_dim : int
+        Attentionの次元数
+    emb_dim : int
+        埋め込み次元数
+    micro_batch : int, default = 1
+        何バッチ毎に進めるか?
+    """
+
+    model_type = "custom_gpt2"
+
+    def __init__(self, vocab_size: int = 50304, max_context_len: int = 1024, layer_num: int = 12, head_num: int = 12, emb_dim: int = 768, micro_batch: int = 1, **kwargs) -> None:
+        super().__init__(**kwargs)
+        self.vocab_size: int = vocab_size
+        self.max_context_len: int = max_context_len
+        self.layer_num: int = layer_num
+        self.head_num: int = head_num
+        self.emb_dim: int = emb_dim
+        self.micro_batch: int = micro_batch
+
+
+class CasualSelfAttention(nn.Module):
+    def __init__(self, config: GPTConfig) -> None:
+        super().__init__()
+        assert config.emb_dim % config.head_num == 0, (
+            f"埋め込み次元数がヘッドの数で割り切れません!"
+        )
+        # すべてのヘッドに対するkey, query, valueの投影を、バッチ処理で行う
+        # (B, T, C) -> (B, T, 3C)
+        self.c_attn = nn.Linear(config.emb_dim, 3 * config.emb_dim)
+        # output projection
+        self.c_proj = nn.Linear(config.emb_dim, config.emb_dim)
+        self.c_proj.NANOGPT_SCALE_INIT = 1  # type: ignore
+        # 正規化
+        self.head_num = config.head_num
+        self.emb_dim = config.emb_dim
+        # Attention Mask
+        self.register_buffer(
+            "mask",
+            torch.tril(torch.ones(config.max_context_len, config.max_context_len)).view(
+                1, 1, config.max_context_len, config.max_context_len
+            ),
+        )
+
+    def forward(self, x: torch.Tensor, use_cache=None, attention_mask=None, **kwargs) -> torch.Tensor:
+        B, T, C = x.size()
+        qkv: torch.Tensor = self.c_attn(x)
+        q, k, v = qkv.split(self.emb_dim, dim=2)
+        # 軽量化のためにreshapeではなくviewを使用
+        # 次元を入れ替え(B, head_num, T, head_dim)
+        # Bとhead_numをバッチとして扱う
+        k = k.view(B, T, self.head_num, C // self.head_num).transpose(1, 2)
+        q = q.view(B, T, self.head_num, C // self.head_num).transpose(1, 2)
+        v = v.view(B, T, self.head_num, C // self.head_num).transpose(1, 2)
+
+        # Attention
+        # attn = (q @ k.transpose(-2, -1)) * (k.size(-1) ** -1)
+        # attn = attn.masked_fill(self.mask[:, :, :T, :T] == 0, float("-inf"))  # type: ignore
+        # attn = F.softmax(attn, dim=-1)
+        # y = attn @ v
+
+        with sdpa_kernel(
+            [
+                SDPBackend.FLASH_ATTENTION,
+                SDPBackend.CUDNN_ATTENTION,
+                SDPBackend.EFFICIENT_ATTENTION,
+                SDPBackend.MATH,
+                SDPBackend.ERROR,
+            ]
+        ):
+            y = F.scaled_dot_product_attention(q, k, v, is_causal=True)
+
+        # (B, T, head_num, head_dim) -> (B, head_num, T, head_dim) -> 各headを結合 -> (B, T, C)
+        y = y.transpose(1, 2).contiguous().view(B, T, C)
+        y = self.c_proj(y)
+
+        return y
+
+
+class MLP(nn.Module):
+    def __init__(self, config: GPTConfig) -> None:
+        super().__init__()
+
+        self.c_fc: nn.Linear = nn.Linear(config.emb_dim, 4 * config.emb_dim)
+        self.gelu = nn.GELU(approximate="tanh")
+        self.c_proj: nn.Linear = nn.Linear(4 * config.emb_dim, config.emb_dim)
+        self.c_proj.NANOGPT_SCALE_INIT = 1  # type: ignore
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = self.c_fc(x)
+        x = self.gelu(x)
+        x = self.c_proj(x)
+        return x
+
+
+class Block(nn.Module):
+    def __init__(self, config: GPTConfig) -> None:
+        super().__init__()
+        # 埋め込み次元を平均0, 分散1に正規化して学習を安定化させる
+        self.ln_1 = nn.LayerNorm(config.emb_dim)
+        self.attn = CasualSelfAttention(config)
+        # 埋め込み次元を平均0, 分散1に正規化して学習を安定化させる
+        self.ln_2 = nn.LayerNorm(config.emb_dim)
+        self.mlp = MLP(config)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        # Attention is all you needではAttention後に正規化を行っていたが、GPT-2ではPre Normに変更
+        x = x + self.attn(self.ln_1(x))
+        x = x + self.mlp(self.ln_2(x))
+        return x
+
+
+class GPT(PreTrainedModel, GenerationMixin):
+    config_class = GPTConfig
+    
+    # HFにWeight tyingを宣言する
+    _tied_weights_keys = {"lm_head.weight": "transformer.wte.weight"}
+    
+    def __init__(self, config: GPTConfig) -> None:
+        super().__init__(config)
+        self.config: GPTConfig = config
+
+        
+
+        self.transformer = nn.ModuleDict(
+            dict(
+                wte=nn.Embedding(config.vocab_size, config.emb_dim),
+                wpe=nn.Embedding(config.max_context_len, config.emb_dim),
+                h=nn.ModuleList([Block(config) for _ in range(config.layer_num)]),
+                ln_f=nn.LayerNorm(config.emb_dim),
+            )
+        )
+        self.lm_head = nn.Linear(config.emb_dim, config.vocab_size, bias=False)
+
+        # 重みを共有する
+        self.transformer.wte.weight = self.lm_head.weight  # type: ignore
+
+        self.apply(self._init_weight)
+
+    def _init_weight(self, module: nn.Module) -> None:
+        """
+        重みの初期化を行う
+        """
+        if isinstance(module, nn.Linear):
+            std = 0.02
+            if hasattr(module, "NANOGPT_SCALE_INIT"):
+                std *= (2 * self.config.layer_num) ** -0.5
+            torch.nn.init.normal_(module.weight, mean=0.0, std=std)
+            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=0.2)
+
+    def forward(
+        self, input_ids: torch.Tensor, labels: torch.Tensor | None = None
+    ) -> CausalLMOutput:
+        B, T = input_ids.size()
+        assert T <= self.config.max_context_len, f"最大コンテキスト長をオーバー!"
+
+        position: torch.Tensor = torch.arange(
+            0, T, dtype=torch.long, device=input_ids.device
+        )
+        position_emb: torch.Tensor = self.transformer.wpe(position)  # type: ignore
+        token_emb: torch.Tensor = self.transformer.wte(input_ids)  # type: ignore
+
+        x: torch.Tensor = token_emb + position_emb
+
+        block: Block
+        for block in self.transformer.h:  # type: ignore
+            x = block(x)
+        x = self.transformer.ln_f(x)  # type: ignore
+
+        # (B, T, vocab_size)
+        logits: torch.Tensor = self.lm_head(x)
+
+        # トレーニング用
+        loss = None
+        if labels is not None:
+            # view(-1, SOME)はSOME行になるように列を自動で調節する
+            # (B, T, vocab_size) -> (B * T, vocab_size)
+            loss = F.cross_entropy(logits.view(-1, logits.size(-1)), labels.view(-1))
+
+        return CausalLMOutput(loss, logits)  # type: ignore
+
+    def configure_optimizer(self, weight_decay: float, device: str) -> optim.AdamW:
+        # 学習が必要なパラメーターをまとめる
+        param_dict = {
+            parameter_name: parameter
+            for parameter_name, parameter in self.named_parameters()
+        }
+        param_dict = {
+            parameter_name: parameter
+            for parameter_name, parameter in param_dict.items()
+            if parameter.requires_grad
+        }
+        # 最適化グループを作成する
+        # 2次元であるパラメーターのみ対象
+        decay_params = [p for n, p in param_dict.items() if p.dim() >= 2]
+        no_decay_params = [p for n, p in param_dict.items() if p.dim() < 2]
+        optim_group = [
+            {"params": decay_params, "weight_decay": weight_decay},
+            {"params": no_decay_params, "weight_decay": 0.0},
+        ]
+
+        num_decay_params = sum(p.numel() for p in decay_params)
+        num_no_decay_params = sum(p.numel() for p in no_decay_params)
+
+        print(
+            f"減衰を行うテンソル: {len(decay_params)}, with {num_decay_params:,}パラメーター"
+        )
+        print(
+            f"減衰を行わないテンソル: {len(no_decay_params)}, with {num_no_decay_params:,}パラメーター"
+        )
+
+        # AdamWを使う
+        is_fused = "fused" in inspect.signature(torch.optim.AdamW).parameters
+        can_use_fused = is_fused and "cuda" in device
+        print("Fused AdamWを使用中") if can_use_fused else print(
+            "Fused AdamWは使用できません"
+        )
+
+        optimizer = optim.AdamW(
+            optim_group, lr=3e-4, betas=(0.9, 0.95), eps=1e-8, fused=can_use_fused
+        )
+
+        return optimizer