Closed MichelNivard closed 1 year ago
I cant use TORCH_STREAM, but I can select it via inquirer I think (but thats a very uneducated guess) its because TORCH_STREAM is not defined in the init.py in /agnostic/backends:
Backends: Dict[str, module] = { "tensorflow(cpu/gpu)": RWKVTFOps, "pytorch(cpu/gpu)": RWKVCudaOps, "numpy(cpu)": RWKVNumpyOps, "jax(cpu/gpu/tpu)": RWKVJaxOps, "pytorch-deepspeed(gpu)": RWKVCudaDeepspeedOps, "pytorch-quant(gpu-8bit)": RWKVCudaQuantOps, "pytorch-stream(gpu-config-vram)": RWKVStreamBigOps, "pytorch-split(2xgpu)": RWKVSplitCudaOps, "export-torchscript": RWKVPTTSExportOps, "export-tensorflow": RWKVTFExport, "onnx_export": RWKVOnnxOps, # "coreml": RWKVCoreMLOps } TORCH = "pytorch(cpu/gpu)" JAX = "jax(cpu/gpu/tpu)" TF = "tensorflow(cpu/gpu)" TORCH_DEEPSPEED = "pytorch-deepspeed(gpu)" TORCH_QUANT = "pytorch-quant(gpu-8bit)" TORCH_SPLIT = "pytorch-split(2xgpu)" TORCH_EXPORT = "export-torchscript" TF_EXPORT = "export-tensorflow" ONNX_EXPORT = "onnx_export" NUMPY = "numpy(cpu)" # COREML = "coreml"
I cant use TORCH_STREAM, but I can select it via inquirer I think (but thats a very uneducated guess) its because TORCH_STREAM is not defined in the init.py in /agnostic/backends: