Merge pull request #293 from modelscope/dev

hunyuanvideo quantization
This commit is contained in:
Zhongjie Duan
2024-12-19 16:20:35 +08:00
committed by GitHub
3 changed files with 3 additions and 3 deletions

View File

@@ -11,7 +11,7 @@ model_manager.load_models(
[
"models/HunyuanVideo/transformers/mp_rank_00_model_states.pt"
],
torch_dtype=torch.bfloat16,
torch_dtype=torch.bfloat16, # you can use torch_dtype=torch.float8_e4m3fn to enable quantization.
device="cpu"
)

View File

@@ -11,7 +11,7 @@ model_manager.load_models(
[
"models/HunyuanVideo/transformers/mp_rank_00_model_states.pt"
],
torch_dtype=torch.bfloat16,
torch_dtype=torch.bfloat16, # you can use torch_dtype=torch.float8_e4m3fn to enable quantization.
device="cpu"
)

View File

@@ -11,7 +11,7 @@ model_manager.load_models(
[
"models/HunyuanVideo/transformers/mp_rank_00_model_states.pt"
],
torch_dtype=torch.bfloat16,
torch_dtype=torch.bfloat16, # you can use torch_dtype=torch.float8_e4m3fn to enable quantization.
device="cuda"
)