Keep last 3 benchmark shapes for fp8_quant and causal_conv1d (#129)

* Keep last 3 benchmark shapes for fp8_quant_py instead of first 3

The larger shapes (higher num_tokens and hidden_dim) are more
representative of production workloads for FP8 quantization.

* Keep last 3 benchmark shapes for causal_conv1d_py instead of first 3

The larger shapes (higher D and S) are more representative of
production workloads for causal conv1d.

Author: yf225

problems/helion/causal_conv1d_py/task.yml

@@ -39,9 +39,9 @@ tests:
   - {"B": 4, "D": 64, "S": 128, "W": 4, "seed": 9173}
 
 benchmarks:
-  - {"B": 1, "D": 768, "S": 512, "W": 4, "seed": 31232}
-  - {"B": 1, "D": 768, "S": 2048, "W": 4, "seed": 4052}
   - {"B": 1, "D": 1536, "S": 2048, "W": 4, "seed": 2146}
+  - {"B": 1, "D": 2560, "S": 2048, "W": 4, "seed": 3129}
+  - {"B": 1, "D": 2560, "S": 4096, "W": 4, "seed": 54352}
 
 test_timeout: 180
 benchmark_timeout: 180

problems/helion/fp8_quant_py/task.yml

@@ -46,9 +46,9 @@ tests:
   - {"num_tokens": 8, "hidden_dim": 4096, "group_size": 128, "seed": 9173}
 
 benchmarks:
-  - {"num_tokens": 1, "hidden_dim": 4096, "group_size": 128, "seed": 31232}
-  - {"num_tokens": 16, "hidden_dim": 4096, "group_size": 128, "seed": 4052}
   - {"num_tokens": 256, "hidden_dim": 4096, "group_size": 128, "seed": 2146}
+  - {"num_tokens": 256, "hidden_dim": 8192, "group_size": 128, "seed": 3129}
+  - {"num_tokens": 4096, "hidden_dim": 7168, "group_size": 128, "seed": 54352}
 
 test_timeout: 180
 benchmark_timeout: 180