biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix(minicpmv): fix audio inference by handling meta device in init_re… (#36751)

Browse Source 
Signed-off-by: caitianchi <caitianchi@modelbest.cn>
This commit is contained in:
tc-mb
2026-03-11 18:06:28 +08:00
committed by GitHub
parent 545d18d81b
commit 4286cc5ec2
1 changed files with 10 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
vllm/model_executor/models/minicpmv.py
View File

@@ -1453,10 +1453,11 @@ class MiniCPMV2_6(MiniCPMVBaseModel, SupportsLoRA):
quant_config=quant_config,
prefix=prefix,
)
return resampler.to(
device=current_platform.device_type, dtype=torch.get_default_dtype()
)
target_device = current_platform.device_type
target_dtype = torch.get_default_dtype()
if any(p.is_meta for p in resampler.parameters()):
return resampler.to_empty(device=target_device).to(dtype=target_dtype)
return resampler.to(device=target_device, dtype=target_dtype)
def get_vision_hidden_states(self, data: MiniCPMVImagePixelInputs) -> torch.Tensor:
pixel_values = data["pixel_values"]
@@ -1649,10 +1650,11 @@ class MiniCPMV4_5(MiniCPMVBaseModel, SupportsLoRA):
quant_config=quant_config,
prefix=prefix,
)
return resampler.to(
device=current_platform.device_type, dtype=torch.get_default_dtype()
)
target_device = current_platform.device_type
target_dtype = torch.get_default_dtype()
if any(p.is_meta for p in resampler.parameters()):
return resampler.to_empty(device=target_device).to(dtype=target_dtype)
return resampler.to(device=target_device, dtype=target_dtype)
def get_vision_hidden_states(self, data: MiniCPMVImagePixelInputs) -> torch.Tensor:
pixel_values = data["pixel_values"]