[None][test] Add tests for all database configs. #11653

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged

fsaady merged 21 commits into NVIDIA:main from fsaady:fsaady/fix-curated-configs

Feb 26, 2026

examples/configs/curated/deepseek-r1-deepgemm.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -6,7 +6,6 @@ trust_remote_code: true @@
     enable_attention_dp: true
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
     kv_cache_config:
       dtype: fp8
       free_gpu_memory_fraction: 0.8
@@ Expand Down @@

examples/configs/curated/deepseek-r1-latency.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -3,10 +3,10 @@ tensor_parallel_size: 8 @@
     moe_expert_parallel_size: 2
     max_num_tokens: 32768
     trust_remote_code: true
-    moe_backend: TRTLLM
-    use_cuda_graph: true
     kv_cache_config:
       free_gpu_memory_fraction: 0.75
+    moe_config:
+      backend: TRTLLM
     speculative_config:
       decoding_type: MTP
       num_nextn_predict_layers: 3
@@ Expand Down @@

examples/configs/curated/deepseek-r1-throughput.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -6,7 +6,6 @@ trust_remote_code: true @@
     enable_attention_dp: true
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
     kv_cache_config:
       dtype: fp8
       free_gpu_memory_fraction: 0.8
@@ Expand Down @@

examples/configs/curated/gpt-oss-120b-latency.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -3,9 +3,6 @@ max_num_tokens: 16384 @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 1
     trust_remote_code: true
-    enable_attention_dp: false
-    kv_cache_config:
-      free_gpu_memory_fraction: 0.9
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 64
@@ Expand Down @@

examples/configs/curated/gpt-oss-120b-throughput.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -4,8 +4,6 @@ tensor_parallel_size: 2 @@
     moe_expert_parallel_size: 2
     trust_remote_code: true
     enable_attention_dp: true
-    kv_cache_config:
-      free_gpu_memory_fraction: 0.9
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 720
@@ Expand Down @@

examples/configs/curated/kimi-k2-thinking.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -4,7 +4,6 @@ max_seq_len: 8212 @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     enable_attention_dp: true
-    pipeline_parallel_size: 1
     print_iter_log: true
     kv_cache_config:
       free_gpu_memory_fraction: 0.75
@@ Expand Down @@

examples/configs/curated/llama-3.3-70b.yaml

-Original file line number
+Diff line change
@@ -1,12 +1,9 @@
     max_batch_size: 1024
     max_num_tokens: 2048
-    tensor_parallel_size: 1
     moe_expert_parallel_size: 1
     trust_remote_code: true
-    enable_attention_dp: false
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 1024
     kv_cache_config:
       dtype: fp8
-      free_gpu_memory_fraction: 0.9

examples/configs/curated/llama-4-scout.yaml

-Original file line number
+Diff line change
@@ -1,12 +1,9 @@
     max_batch_size: 1024
     max_num_tokens: 2048
-    tensor_parallel_size: 1
     moe_expert_parallel_size: 1
     trust_remote_code: true
-    enable_attention_dp: false
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 1024
     kv_cache_config:
       dtype: fp8
-      free_gpu_memory_fraction: 0.9

examples/configs/curated/lookup.yaml

-Original file line number
+Diff line change
@@ -0,0 +1,34 @@
+    # arch: MODEL_CLASS_MAPPING key; required when model has get_model_defaults. Add when adding entries.
+    - model: Qwen/Qwen3-Next-80B-A3B-Thinking
+      arch: Qwen3NextForCausalLM
+      config_path: examples/configs/curated/qwen3-next.yaml
+    - model: Qwen/Qwen3-30B-A3B
+      arch: Qwen3MoeForCausalLM
+      config_path: examples/configs/curated/qwen3.yaml
+    - model: Qwen/Qwen3-30B-A3B
+      arch: Qwen3MoeForCausalLM
+      config_path: examples/configs/curated/qwen3-disagg-prefill.yaml
+    - model: deepseek-ai/DeepSeek-R1-0528
+      arch: DeepseekV3ForCausalLM
+      config_path: examples/configs/curated/deepseek-r1-latency.yaml
+    - model: deepseek-ai/DeepSeek-R1-0528
+      arch: DeepseekV3ForCausalLM
+      config_path: examples/configs/curated/deepseek-r1-throughput.yaml
+    - model: deepseek-ai/DeepSeek-R1-0528
+      arch: DeepseekV3ForCausalLM
+      config_path: examples/configs/curated/deepseek-r1-deepgemm.yaml
+    - model: openai/gpt-oss-120b
+      arch: GptOssForCausalLM
+      config_path: examples/configs/curated/gpt-oss-120b-latency.yaml
+    - model: openai/gpt-oss-120b
+      arch: GptOssForCausalLM
+      config_path: examples/configs/curated/gpt-oss-120b-throughput.yaml
+    - model: nvidia/Llama-3.3-70B-Instruct-FP8
+      arch: LlamaForCausalLM
+      config_path: examples/configs/curated/llama-3.3-70b.yaml
+    - model: nvidia/Llama-4-Scout-17B-16E-Instruct-FP8
+      arch: Llama4ForConditionalGeneration
+      config_path: examples/configs/curated/llama-4-scout.yaml
+    - model: nvidia/Kimi-K2-Thinking-NVFP4
+      arch: DeepseekV3ForCausalLM
+      config_path: examples/configs/curated/kimi-k2-thinking.yaml

examples/configs/curated/qwen3-disagg-prefill.yaml

-Original file line number
+Diff line change
@@ -1,9 +1,7 @@
     max_batch_size: 161
     max_num_tokens: 1160
-    kv_cache_config:
-      free_gpu_memory_fraction: 0.8
-    tensor_parallel_size: 1
     moe_expert_parallel_size: 1
     trust_remote_code: true
-    print_iter_log: true
     enable_attention_dp: true
+    kv_cache_config:
+      free_gpu_memory_fraction: 0.8

examples/configs/curated/qwen3-next.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -3,14 +3,10 @@ max_num_tokens: 4096 @@
     tensor_parallel_size: 4
     moe_expert_parallel_size: 4
     trust_remote_code: true
-    enable_attention_dp: false
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 720
     moe_config:
-        backend: TRTLLM
+      backend: TRTLLM
     stream_interval: 20
     num_postprocess_workers: 4
-    kv_cache_config:
-      enable_block_reuse: false
-      free_gpu_memory_fraction: 0.9

examples/configs/curated/qwen3.yaml

-Original file line number
+Diff line change
@@ -1,8 +1,5 @@
     max_batch_size: 161
     max_num_tokens: 1160
-    kv_cache_config:
-      free_gpu_memory_fraction: 0.8
-    tensor_parallel_size: 1
     moe_expert_parallel_size: 1
     cuda_graph_config:
       enable_padding: true
@@ Expand All / @@ -17,5 +14,6 @@ cuda_graph_config: @@
       - 128
       - 256
       - 384
-    print_iter_log: true
     enable_attention_dp: true
+    kv_cache_config:
+      free_gpu_memory_fraction: 0.8

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/1k1k_tp8_conc16.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/1k1k_tp8_conc32.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/1k1k_tp8_conc4.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/1k1k_tp8_conc64.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/1k1k_tp8_conc8.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/8k1k_tp8_conc16.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/8k1k_tp8_conc32.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/8k1k_tp8_conc4.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/8k1k_tp8_conc64.yaml

-Original file line number
+Diff line change
@@ Expand Up / @@ -16,6 +16,5 @@ attention_dp_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/B200/8k1k_tp8_conc8.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,6 @@
     cuda_graph_config:
       enable_padding: true
       max_batch_size: 256
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +11,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/1k1k_tp8_conc16.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/1k1k_tp8_conc32.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/1k1k_tp8_conc4.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/1k1k_tp8_conc64.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/1k1k_tp8_conc8.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 1152
     max_seq_len: 2068

examples/configs/database/deepseek-ai/DeepSeek-R1-0528/H200/8k1k_tp8_conc16.yaml

-Original file line number
+Diff line change
@@ -1,7 +1,5 @@
     cuda_graph_config:
       enable_padding: true
-      max_batch_size: 128
-    enable_attention_dp: false
     print_iter_log: true
     kv_cache_config:
       dtype: fp8
@@ Expand All / @@ -12,6 +10,5 @@ moe_config: @@
     tensor_parallel_size: 8
     moe_expert_parallel_size: 8
     trust_remote_code: true
-    backend: pytorch
     max_num_tokens: 8320
     max_seq_len: 9416

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[None][test] Add tests for all database configs. #11653

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!