Jack Zhou

### PR types New features ### PR changes Others ### Description Use fastdeploy instead of inference backend

PNDMScheduler has a considerable cost when running StableDiffusionPipeline in diffuser-0.4.1

1

### Describe the bug ## Enviroment GPU: A10, CUDA 11.6, cuDNN 8.4.0 Torch: 1.12.1 diffuser: 0.4.1 ## Phenomenon When I ran the StableDiffusionPipeline with fp16 precision, I found the time...

### PR types New features ### PR changes APIs ### Describe Add collect shape for pp-trt backend

### PR types New features ### PR changes Others ### Describe Add stable diffusion model based on fastdeploy

[Model] Add text classification task for ernie-3.0

1

### PR types(PR类型) Model ### Describe - Add text classification example for ernie-3.0

### PR types(PR类型) Benchmark ### Describe Add benchmark for ernie sequence classification. Statistics include: - Time cost of each stage, including tokenization, runtime and postprocessing - CPU memory cost -...

### PR types New features ### PR changes APIs ### Description Add ByteLevel pretokenizer and RobertaProcessing

### PR types(PR类型) Benchmark ### Describe Add uie benchmark

### PR types(PR类型) Diffusion ### Describe Add C++ dpm solver

### PR types(PR类型) Backend ### Describe - Add delete pass python api - Add DisablePaddleTrtOPs