Update skyrl/backends/skyrl_train/weight_sync/fp8/quantize.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: jinghanyao1-hub

skyrl/backends/skyrl_train/weight_sync/fp8/quantize.py

@@ -72,7 +72,7 @@ def blockwise_cast_to_fp8(
     padded_cols = ((cols + block_n - 1) // block_n) * block_n
 
     fp8_info = torch.finfo(torch.float8_e4m3fn)
-    weight_fp32 = weight.detach().to(torch.float32)
+    weight_fp32 = weight.detach().to(torch.float32).contiguous()
     if padded_rows != rows or padded_cols != cols:
         padded = weight_fp32.new_zeros((padded_rows, padded_cols))
         padded[:rows, :cols].copy_(weight_fp32)