File size: 10,904 Bytes
51115cb | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 | """BananaMind 2 Nano implementation for Hugging Face Transformers."""
import math
from typing import Optional
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import PreTrainedModel
from transformers.cache_utils import Cache, DynamicCache
from transformers.generation.utils import GenerationMixin
from transformers.modeling_outputs import CausalLMOutputWithPast
from .configuration_bananamind2nano import BananaMind2NanoConfig
class RMSNorm(nn.Module):
def __init__(self, dim, eps=1e-6):
super().__init__()
self.eps = eps
self.weight = nn.Parameter(torch.ones(dim))
def forward(self, x):
x_float = x.float()
rms = torch.rsqrt(x_float.pow(2).mean(-1, keepdim=True) + self.eps)
return (x_float * rms * self.weight.float()).type_as(x)
def build_rope_inv_freq(head_dim, theta=100000.0):
return 1.0 / (theta ** (torch.arange(0, head_dim, 2, dtype=torch.float32) / head_dim))
def precompute_freqs_cis(head_dim, seq_len, theta=100000.0):
freqs = build_rope_inv_freq(head_dim, theta)
positions = torch.arange(seq_len, dtype=torch.float32)
freqs = torch.outer(positions, freqs)
return torch.polar(torch.ones_like(freqs), freqs)
def apply_rotary_emb(q, k, freqs_cis):
q_complex = torch.view_as_complex(q.float().reshape(*q.shape[:-1], -1, 2))
k_complex = torch.view_as_complex(k.float().reshape(*k.shape[:-1], -1, 2))
freqs_cis = freqs_cis.unsqueeze(0).unsqueeze(0)
q_out = torch.view_as_real(q_complex * freqs_cis).flatten(-2)
k_out = torch.view_as_real(k_complex * freqs_cis).flatten(-2)
return q_out.type_as(q), k_out.type_as(k)
class BananaMind2NanoAttention(nn.Module):
def __init__(self, config, layer_idx):
super().__init__()
self.layer_idx = layer_idx
self.n_head = config.num_attention_heads
self.n_kv_heads = config.num_key_value_heads
self.head_dim = config.head_dim
self.n_rep = self.n_head // self.n_kv_heads
self.q_proj = nn.Linear(config.hidden_size, self.n_head * self.head_dim, bias=False)
self.k_proj = nn.Linear(config.hidden_size, self.n_kv_heads * self.head_dim, bias=False)
self.v_proj = nn.Linear(config.hidden_size, self.n_kv_heads * self.head_dim, bias=False)
self.o_proj = nn.Linear(self.n_head * self.head_dim, config.hidden_size, bias=False)
self.q_norm = RMSNorm(self.head_dim, eps=config.rms_norm_eps)
self.k_norm = RMSNorm(self.head_dim, eps=config.rms_norm_eps)
def forward(
self,
x,
freqs_cis,
attention_mask=None,
past_key_values=None,
use_cache=False,
):
batch_size, seq_len, _ = x.shape
q = self.q_proj(x).view(
batch_size,
seq_len,
self.n_head,
self.head_dim,
).transpose(1, 2)
k = self.k_proj(x).view(
batch_size,
seq_len,
self.n_kv_heads,
self.head_dim,
).transpose(1, 2)
v = self.v_proj(x).view(
batch_size,
seq_len,
self.n_kv_heads,
self.head_dim,
).transpose(1, 2)
q = self.q_norm(q)
k = self.k_norm(k)
q, k = apply_rotary_emb(q, k, freqs_cis)
past_length = 0
if use_cache and past_key_values is not None:
past_length = past_key_values.get_seq_length(self.layer_idx)
k, v = past_key_values.update(k, v, self.layer_idx)
kv_len = k.size(-2)
k = k.unsqueeze(2).expand(
batch_size,
self.n_kv_heads,
self.n_rep,
kv_len,
self.head_dim,
).reshape(batch_size, self.n_head, kv_len, self.head_dim)
v = v.unsqueeze(2).expand(
batch_size,
self.n_kv_heads,
self.n_rep,
kv_len,
self.head_dim,
).reshape(batch_size, self.n_head, kv_len, self.head_dim)
attn_mask = None
is_causal = past_length == 0 and attention_mask is None
if not is_causal:
query_positions = past_length + torch.arange(seq_len, device=x.device)
key_positions = torch.arange(kv_len, device=x.device)
causal = key_positions.unsqueeze(0) <= query_positions.unsqueeze(1)
attn_mask = causal[None, None, :, :]
if attention_mask is not None:
key_padding = attention_mask.to(torch.bool)
if key_padding.size(-1) < kv_len:
cached_padding = torch.ones(
key_padding.size(0),
kv_len - key_padding.size(-1),
dtype=torch.bool,
device=key_padding.device,
)
key_padding = torch.cat((cached_padding, key_padding), dim=-1)
else:
key_padding = key_padding[:, -kv_len:]
attn_mask = attn_mask & key_padding[:, None, None, :]
is_causal = False
y = F.scaled_dot_product_attention(
q,
k,
v,
attn_mask=attn_mask,
is_causal=is_causal,
)
y = y.transpose(1, 2).contiguous().view(
batch_size,
seq_len,
self.n_head * self.head_dim,
)
return self.o_proj(y)
class BananaMind2NanoSwiGLUMLP(nn.Module):
def __init__(self, config):
super().__init__()
self.w_gate = nn.Linear(config.hidden_size, config.intermediate_size, bias=False)
self.w_up = nn.Linear(config.hidden_size, config.intermediate_size, bias=False)
self.w_down = nn.Linear(config.intermediate_size, config.hidden_size, bias=False)
def forward(self, x):
return self.w_down(F.silu(self.w_gate(x)) * self.w_up(x))
class BananaMind2NanoBlock(nn.Module):
def __init__(self, config, layer_idx):
super().__init__()
self.ln_1 = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.attn = BananaMind2NanoAttention(config, layer_idx)
self.ln_2 = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
self.mlp = BananaMind2NanoSwiGLUMLP(config)
def forward(
self,
x,
freqs_cis,
attention_mask=None,
past_key_values=None,
use_cache=False,
):
x = x + self.attn(
self.ln_1(x),
freqs_cis,
attention_mask=attention_mask,
past_key_values=past_key_values,
use_cache=use_cache,
)
return x + self.mlp(self.ln_2(x))
class BananaMind2NanoPreTrainedModel(PreTrainedModel):
config_class = BananaMind2NanoConfig
base_model_prefix = "transformer"
supports_gradient_checkpointing = False
def _init_weights(self, module):
std = 0.02
if hasattr(module, "NANOGPT_SCALE_INIT"):
std *= 2 * self.config.num_hidden_layers ** -0.5
if isinstance(module, nn.Linear):
nn.init.normal_(module.weight, mean=0.0, std=std)
elif isinstance(module, nn.Embedding):
nn.init.normal_(module.weight, mean=0.0, std=0.02)
class BananaMind2NanoForCausalLM(BananaMind2NanoPreTrainedModel, GenerationMixin):
_tied_weights_keys = {"lm_head.weight": "transformer.wte.weight"}
def __init__(self, config):
super().__init__(config)
self.config = config
self.transformer = nn.ModuleDict(
{
"wte": nn.Embedding(config.vocab_size, config.hidden_size),
"h": nn.ModuleList(
[
BananaMind2NanoBlock(config, layer_idx)
for layer_idx in range(config.num_hidden_layers)
]
),
"ln_f": RMSNorm(config.hidden_size, eps=config.rms_norm_eps),
}
)
self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
if config.tie_word_embeddings:
self.lm_head.weight = self.transformer["wte"].weight
self._embd_scale = math.sqrt(config.hidden_size)
self._freqs_cis_cache = None
self.post_init()
def get_input_embeddings(self):
return self.transformer["wte"]
def set_input_embeddings(self, value):
self.transformer["wte"] = value
def get_output_embeddings(self):
return self.lm_head
def set_output_embeddings(self, new_embeddings):
self.lm_head = new_embeddings
def _get_freqs_cis(self, seq_len, device):
cache = self._freqs_cis_cache
if cache is None or cache.device != device or cache.size(0) < seq_len:
cache = precompute_freqs_cis(
self.config.head_dim,
seq_len,
self.config.rope_theta,
).to(device)
self._freqs_cis_cache = cache
return cache[:seq_len]
def forward(
self,
input_ids,
attention_mask=None,
labels=None,
past_key_values: Optional[Cache] = None,
use_cache=None,
**kwargs,
):
_, seq_len = input_ids.shape
if use_cache is None:
use_cache = self.config.use_cache and labels is None
if use_cache and past_key_values is None:
past_key_values = DynamicCache(config=self.config)
past_length = past_key_values.get_seq_length() if use_cache else 0
total_length = past_length + seq_len
if total_length > self.config.max_position_embeddings:
raise ValueError(
f"Sequence length {total_length} exceeds the configured maximum "
f"of {self.config.max_position_embeddings}"
)
x = self.transformer["wte"](input_ids) * self._embd_scale
freqs_cis = self._get_freqs_cis(total_length, input_ids.device)[past_length:]
for block in self.transformer["h"]:
x = block(
x,
freqs_cis,
attention_mask=attention_mask,
past_key_values=past_key_values,
use_cache=use_cache,
)
x = self.transformer["ln_f"](x)
logits = self.lm_head(x)
loss = None
if labels is not None:
shift_logits = logits[..., :-1, :].contiguous()
shift_labels = labels[..., 1:].contiguous()
loss = F.cross_entropy(
shift_logits.float().reshape(-1, shift_logits.size(-1)),
shift_labels.reshape(-1),
)
if self.config.z_loss_coeff:
loss = loss + self.config.z_loss_coeff * logits.float().pow(2).mean()
return CausalLMOutputWithPast(
loss=loss,
logits=logits,
past_key_values=past_key_values if use_cache else None,
)
|