Revert "[Renderer] Separate out RendererConfig from ModelConfig (#30145)" (#30199)

This commit is contained in:
Cyrus Leung
2025-12-07 16:00:22 +08:00
committed by GitHub
parent 27f4c2fd46
commit e83b7e379c
105 changed files with 797 additions and 969 deletions

View File

@@ -8,7 +8,7 @@ import torch
from safetensors.torch import load_file
from torch import nn
from vllm.config import ModelConfig, RendererConfig, VllmConfig
from vllm.config import ModelConfig, VllmConfig
from vllm.config.lora import LoRAConfig
from vllm.lora.layers import (
ColumnParallelLinearWithLoRA,
@@ -422,11 +422,7 @@ def test_lru_cache_worker_adapter_manager(dist_init, dummy_model, device, tmp_pa
)
model_config = ModelConfig(max_model_len=16)
vllm_config = VllmConfig(
model_config=model_config,
renderer_config=RendererConfig(model_config=model_config),
lora_config=lora_config,
)
vllm_config = VllmConfig(model_config=model_config, lora_config=lora_config)
vllm_config.scheduler_config.max_num_seqs = 4
vllm_config.scheduler_config.max_num_batched_tokens = 2
@@ -529,11 +525,7 @@ def test_worker_adapter_manager(dist_init, dummy_model_gate_up, device, tmp_path
)
model_config = ModelConfig(max_model_len=16)
vllm_config = VllmConfig(
model_config=model_config,
renderer_config=RendererConfig(model_config=model_config),
lora_config=lora_config,
)
vllm_config = VllmConfig(model_config=model_config, lora_config=lora_config)
vllm_config.scheduler_config.max_num_seqs = 4
vllm_config.scheduler_config.max_num_batched_tokens = 2