save fp8 disagg test

Author: brb-nv

tests/integration/defs/disaggregated/test_configs/disagg_config_ctxtp2_gentp1cp2_deepseek_v3_lite_fp8_tllm_gen.yaml

@@ -0,0 +1,40 @@
+hostname: localhost
+port: 8000
+model: DeepSeek-V3-Lite/fp8
+free_gpu_memory_fraction: 0.25
+backend: "pytorch"
+disable_overlap_scheduler: True
+cuda_graph_config: null
+context_servers:
+  num_instances: 1
+  enable_chunked_prefill: False
+  kv_cache_config:
+    enable_block_reuse: False
+    enable_partial_reuse: False
+  moe_config:
+    backend: DEEPGEMM
+  tensor_parallel_size: 2
+  pipeline_parallel_size: 1
+  cache_transceiver_config:
+    backend: "UCX"
+    max_tokens_in_buffer: 4096
+  urls:
+      - "localhost:8001"
+generation_servers:
+  num_instances: 1
+  tensor_parallel_size: 1
+  pipeline_parallel_size: 1
+  context_parallel_size: 2
+  cp_config:
+    cp_type: helix
+  enable_chunked_prefill: False
+  kv_cache_config:
+    enable_block_reuse: False
+    enable_partial_reuse: False
+  cache_transceiver_config:
+    backend: "UCX"
+    max_tokens_in_buffer: 4096
+  moe_config:
+    backend: DEEPGEMM
+  urls:
+      - "localhost:8002"

tests/integration/defs/disaggregated/test_disaggregated.py

@@ -271,6 +271,10 @@ def get_test_config(test_desc, example_dir, test_root):
         (4,
          f"{test_configs_root}/disagg_config_ctxtp2_gentp1cp2_deepseek_v3_lite_bf16_tllm_gen.yaml"
          ),
+        "deepseek_v3_lite_fp8_tllm_gen_helix":
+        (4,
+         f"{test_configs_root}/disagg_config_ctxtp2_gentp1cp2_deepseek_v3_lite_fp8_tllm_gen.yaml"
+        ),
     }
 
     if test_desc not in config_map: