[BUG] e4m3, int8, bf16 pytorch emitter not working

I am attempting to emit pytorch code but unfortunately it does not work for fp8, bf16, and int8. I have tried to patch the converter type dict https://github.com/OrenLeung/cutlass/commit/6d619c964eb8b9c150a5f97891849d33f6ee8b64

This patch fixed the initial set of issue but unfortunately I run into a deeper issue when the emitter tries to compile the kernel.

New Error Issue

c /workspace/cutlass/examples/python/gemm_mod_fp8_kernel.cu -o gemm_mod_fp8_kernel.cuda.o 
/workspace/cutlass/examples/python/gemm_mod_fp8_kernel.cu(100): error: a value of type "cutlass::float_e4m3_t *" cannot be used to initialize an entity of type "const float *"
          D,
          ^

Reprod

import torch
import cutlass

# FP8 CUTLASS GEMM plan
plan_fp8 = cutlass.op.Gemm(
    element=torch.float8_e4m3fn,
    element_accumulator=torch.float32,
    element_D=torch.float32,
    layout_A=cutlass.LayoutType.RowMajor,
    layout_B=cutlass.LayoutType.ColumnMajor,
    layout_C=cutlass.LayoutType.ColumnMajor)
op_fp8 = plan_fp8.construct()

# Generate the PyTorch module for the FP8 GEMM operation
mod_fp8 = cutlass.emit.pytorch(op_fp8, name='gemm_mod_fp8', cc=plan_fp8.cc, jit=True)

# BF16 CUTLASS GEMM plan
plan_bf16 = cutlass.op.Gemm(
    element=torch.bfloat16,
    element_accumulator=torch.float32,
    element_D=torch.float32,
    layout_A=cutlass.LayoutType.RowMajor,
    layout_B=cutlass.LayoutType.ColumnMajor,
    layout_C=cutlass.LayoutType.ColumnMajor)
op_bf16 = plan_bf16.construct()

# Generate the PyTorch module for the BF16 GEMM operation
mod_bf16 = cutlass.emit.pytorch(op_bf16, name='gemm_mod_bf16', cc=plan_bf16.cc, jit=True)

plan = cutlass.op.Gemm(
    element=cutlass.DataType.s8,
    element_accumulator=cutlass.DataType.s32,
    element_D=cutlass.DataType.s32,
    layout=cutlass.LayoutType.RowMajor)
op = plan.construct()

# Generate the PyTorch module for the GEMM operation
mod = cutlass.emit.pytorch(op, name='gemm_mod', cc=plan.cc, jit=True)

plan = cutlass.op.Gemm(
    element=cutlass.DataType.s8,
    element_accumulator=cutlass.DataType.s32,
    element_D=cutlass.DataType.s32,
    layout=cutlass.LayoutType.RowMajor)
op = plan.construct()

# Generate the PyTorch module for the GEMM operation
mod = cutlass.emit.pytorch(op, name='gemm_mod', cc=plan.cc, jit=True)

Start with the provided cutlass.emit.pytorch reproducer and inspect the generated examples/python/gemm_mod_fp8_kernel.cu, especially the failing call near line 100. Re-run the FP8, BF16, and int8 cases after investigating the emitter's generated types; done means each case emits and compiles without the reported pointer-type error.