Fix skip softmax defaults

Signed-off-by: Rohan Joshi <rohjoshi@nvidia.com>

Author: rohansjoshi

modelopt/torch/sparsity/attention_sparsity/config.py

@@ -399,9 +399,9 @@ class FlashSkipSoftmaxConfig(SparseAttentionConfig):
 SKIP_SOFTMAX_CALIB = {
     "sparse_cfg": {
         "calibration": {
-            "target_sparse_ratio": {"prefill": 0.9, "decode": 0.9},
+            "target_sparse_ratio": {"prefill": 0.5, "decode": 0.5},
             "samples": 64,
-            "max_seqlen": 65536,
+            "max_seqlen": 16384,
             "chunk_size": 4096,
         },
         "*attn*": {