[5/N] pass the whole config to model (#9983)

Signed-off-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
youkaichao
2024-11-09 14:17:28 +08:00
committed by GitHub
parent 49d2a41a86
commit 1a95f10ee7
75 changed files with 583 additions and 654 deletions
+8 -8
View File
@@ -3,8 +3,7 @@ import re
from array import array
from dataclasses import dataclass
from functools import lru_cache, partial
from typing import (Any, Iterable, List, Mapping, Optional, Tuple, TypedDict,
Union)
from typing import Iterable, List, Mapping, Optional, Tuple, TypedDict, Union
import torch
from einops import rearrange
@@ -16,7 +15,7 @@ from transformers import PretrainedConfig
from vllm.attention import Attention, AttentionMetadata
from vllm.attention.selector import _Backend
from vllm.compilation.decorators import support_torch_compile
from vllm.config import CacheConfig, MultiModalConfig
from vllm.config import CacheConfig, VllmConfig
from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
split_tensor_along_last_dim,
@@ -1027,13 +1026,14 @@ class MolmoForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
def __init__(
self,
config: PretrainedConfig,
multimodal_config: Optional[MultiModalConfig] = None,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[Mapping[str, Any]] = None,
vllm_config: VllmConfig,
prefix: str = "",
) -> None:
super().__init__()
config = vllm_config.model_config.hf_config
cache_config = vllm_config.cache_config
quant_config = vllm_config.quant_config
multimodal_config = vllm_config.model_config.multimodal_config
self.config = config
self.multimodal_config = multimodal_config