ViperEkura
deb2d7e127
- Replace CacheView/ContiguousCache/PageCache with SGLang-inspired design: KVStorage (flat token-level NHD buffers [n_layers, size, H, D]), ReqToTokenPool (index table [req_idx, pos] -> token_slot), Allocator + PrefixCache (slot allocation with LRU and prefix sharing) - Add KVCache as pure dataclass passed to model: k_buffer, v_buffer, req_to_token, req_pool_indices, seq_lens, out_cache_loc - PagePool orchestrates all three layers, supports contiguous mode (pre-allocated per-request blocks, default) and paged mode (page_size=1 or >1 with dynamic allocation and prefix caching) - Attention layers now do raw buffer indexing instead of opaque write/gather method calls on CacheView objects - Update executor.bind_tasks signature: seq_lens list + start_pos - Rename paged_cache -> kv_cache throughout model/ and inference/
50 lines
1.6 KiB
Python
50 lines
1.6 KiB
Python
from dataclasses import asdict
|
|
from typing import Optional
|
|
|
|
import torch.nn as nn
|
|
from torch import Tensor
|
|
|
|
from astrai.inference.core.cache import KVCache
|
|
from astrai.model.components.attention import AttnFactory
|
|
from astrai.model.components.mlp import FFNFactory
|
|
from astrai.model.components.norm import RMSNorm
|
|
|
|
|
|
class DecoderBlock(nn.Module):
|
|
def __init__(self, config, layer_id: int):
|
|
super().__init__()
|
|
cfg = asdict(config)
|
|
cfg.update(
|
|
dim=config.hidden_size,
|
|
dim_ffn=config.intermediate_size,
|
|
n_layers=config.num_hidden_layers,
|
|
n_heads=config.num_attention_heads,
|
|
n_kv_heads=config.num_key_value_heads,
|
|
norm_eps=config.rms_norm_eps,
|
|
down_init_std=0.02 / (2 * config.num_hidden_layers) ** 0.5,
|
|
)
|
|
self.attention = AttnFactory.create(config.attn_type, **cfg, layer_id=layer_id)
|
|
self.input_norm = RMSNorm(config.hidden_size, config.rms_norm_eps)
|
|
self.post_attention_norm = RMSNorm(config.hidden_size, config.rms_norm_eps)
|
|
self.mlp = FFNFactory.create(config.ffn_type, **cfg)
|
|
|
|
def forward(
|
|
self,
|
|
x: Tensor,
|
|
rotary_emb: Tensor,
|
|
attention_mask: Optional[Tensor] = None,
|
|
kv_cache: Optional[KVCache] = None,
|
|
is_causal: bool = False,
|
|
) -> Tensor:
|
|
attn_output = self.attention(
|
|
self.input_norm(x),
|
|
rotary_emb,
|
|
attention_mask,
|
|
kv_cache,
|
|
is_causal,
|
|
)
|
|
x = attn_output + x
|
|
x = self.mlp(self.post_attention_norm(x)) + x
|
|
|
|
return x
|