Chapter 3: Loading the GPT-2 Checkpoint

The from_pretrained class method connects our hand-written GPT implementation to an existing Hugging Face checkpoint. It supports all four GPT-2 sizes by selecting the corresponding number of Transformer blocks, attention heads, and embedding dimensions. The vocabulary size (50257) and context length (1024) stay fixed across these GPT-2 checkpoints.

The method first creates a randomly initialized model with the selected configuration and obtains its state_dict. It then loads the matching Hugging Face model and obtains a second state_dict. The two key lists are compared after removing attention-mask buffers: these masks are implementation details, not learned parameters, so they should not be copied like weights.

class GPT(nn.Module):

    def __init__(self, config):
        super().__init__()
        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),
            h = nn.ModuleList([Block(config) for _ in range(config.n_layer)]),
            ln_f = nn.LayerNorm(config.n_embd),
        ))
        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)

    @classmethod
    def from_pretrained(cls, model_type):
        """Loads pretrained GPT-2 model weights from huggingface"""
        assert model_type in {'gpt2', 'gpt2-medium', 'gpt2-large', 'gpt2-xl'}
        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]
        config_args['vocab_size'] = 50257 # always 50257 for GPT model checkpoints
        config_args['block_size'] = 1024 # always 1024 for GPT model checkpoints
        # 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])
        print("load pretrained model successfully")
        return model

For most tensors, loading is a direct copy. The important exception is the four matrices whose names end in attn.c_attn.weight, attn.c_proj.weight, mlp.c_fc.weight, or mlp.c_proj.weight. Hugging Face represents these layers with a GPT-2-style Conv1D layout, while our implementation uses standard nn.Linear layers. Their dimensions are therefore reversed, so the code transposes the pretrained matrix before copying it.

The shape assertions protect the conversion from silent errors: every ordinary parameter must have the same shape, and every transposed parameter must match after reversing its dimensions. The copies are performed inside torch.no_grad() because loading a checkpoint is initialization, not a differentiable training operation. Once all tensors have been copied, the returned model has the same learned parameters as the selected pretrained GPT-2 checkpoint.

model = GPT.from_pretrained("gpt2")

loading weights from pretrained gpt: gpt2
load pretrained model successfully