[CI][RFC] Replace black-jupyter with ruff-format (#37210)

Co-authored-by: Alison Shao <a.shao@wustl.edu>

benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py

@@ -59,7 +59,6 @@ def tl_gemm(
             bx,
             by,
         ):
-
             A_shared = T.alloc_shared(A_shared_shape, in_dtype)
             B_shared = T.alloc_shared(B_shared_shape, in_dtype)
             C_shared = T.alloc_shared(C_shared_shape, out_dtype)
@@ -350,7 +349,7 @@ def get_benchmark(tp_size):
         tflops = flops / (ms * 1e-3) / 1e12
 
         # Print shape-specific results with TFLOPS
-        print(f"Time: {ms*1000:.2f} ms, TFLOPS: {tflops:.2f}")
+        print(f"Time: {ms * 1000:.2f} ms, TFLOPS: {tflops:.2f}")
         return ms * 1000, max_ms * 1000, min_ms * 1000  # convert to ms
 
     return benchmark

benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm_blackwell.py

@@ -224,7 +224,7 @@ def _benchmark(m, n, k, tp_size, provider):
     tflops = flops / (ms * 1e-3) / 1e12
 
     # Print shape-specific results with TFLOPS
-    print(f"Time: {ms*1000:.2f} us, TFLOPS: {tflops:.2f}")
+    print(f"Time: {ms * 1000:.2f} us, TFLOPS: {tflops:.2f}")
     return ms, max_ms, min_ms
 
 

benchmark/kernels/deepseek/benchmark_deepgemm_fp8_group_gemm.py

@@ -435,7 +435,7 @@ def get_benchmark(tp_size):
         flops = 2 * m * n * k  # multiply-adds
         tflops = flops / (ms * 1e-3) / 1e12
 
-        print(f"Time: {ms*1000:.2f} ms, TFLOPS: {tflops:.2f}")
+        print(f"Time: {ms * 1000:.2f} ms, TFLOPS: {tflops:.2f}")
         return ms * 1000, max_ms * 1000, min_ms * 1000  # convert to ms
 
     return benchmark

benchmark/kernels/deepseek/benchmark_q8kv8_kv_gather.py

@@ -243,8 +243,7 @@ def main():
             else:
                 speedup = f"{legacy_us / us:.2f}x"
             print(
-                f"{s_q:>5} {kv_len:>7} {name:>10} {us:>9.1f} "
-                f"{tbps:>9.3f} {speedup:>8}"
+                f"{s_q:>5} {kv_len:>7} {name:>10} {us:>9.1f} {tbps:>9.3f} {speedup:>8}"
             )
         print()