[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
+7 -5
View File
@@ -9,7 +9,7 @@ from torch import nn
from transformers import ChameleonConfig, ChameleonVQVAEConfig
from vllm.attention import Attention, AttentionMetadata
from vllm.config import CacheConfig, MultiModalConfig
from vllm.config import CacheConfig, VllmConfig
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
from vllm.inputs import (INPUT_REGISTRY, DecoderOnlyInputs, DummyData,
InputContext, token_inputs)
@@ -926,12 +926,14 @@ class ChameleonForConditionalGeneration(nn.Module, SupportsMultiModal,
def __init__(
self,
config: ChameleonConfig,
multimodal_config: MultiModalConfig,
cache_config: Optional[CacheConfig] = None,
quant_config: Optional[QuantizationConfig] = 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
self.model = ChameleonModel(config, cache_config, quant_config)