New

Qwen/Qwen2.5-VL-7B-Instruct/accuracy/tasks.yml

@@ -1,31 +1,5 @@
-# from gs://nm-vllm-certs/model-validation/lmeval/Qwen/Qwen2.5-7B/cuda/0.9.1.dev238+g922878ced/k8s-a100-duo/llm_eval_16218386266.json
 tasks:
-  - name: arc_challenge
-    metrics:
-      - name: acc_norm,none
-        value: 0.6373
-
   - name: gsm8k
     metrics:
       - name: exact_match,strict-match
-        value: 0.8074
-
-  - name: hellaswag
-    metrics:
-      - name: acc_norm,none
-        value: 0.8024
-
-  - name: mmlu
-    metrics:
-      - name: acc,none
-        value: 0.7424
-
-  - name: truthfulqa_mc2
-    metrics:
-      - name: acc,none
-        value: 0.5633
-
-  - name: winogrande
-    metrics:
-      - name: acc,none
-        value: 0.7505
+        value: 0.6944

Qwen/Qwen3-8B-FP8/accuracy/tasks.yml

@@ -0,0 +1,5 @@
+tasks:
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.8756

Qwen/Qwen3-8B-FP8/storage.yml

@@ -0,0 +1,3 @@
+# storage configs for https://huggingface.co/Qwen/Qwen3-8B
+model: hf
+data: hf

RedHatAI/Kimi-K2-Instruct-quantized.w4a16/accuracy/server.yml

@@ -1,3 +1,3 @@
 trust-remote-code: true
-tensor-parallel-size: 2
+tensor-parallel-size: 4
 max-model-len: 16384

RedHatAI/Qwen2.5-VL-7B-Instruct-FP8-Dynamic/accuracy/tasks.yml

@@ -0,0 +1,5 @@
+tasks:
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.6944

RedHatAI/Qwen2.5-VL-7B-Instruct-FP8-Dynamic/storage.yml

@@ -0,0 +1,3 @@
+# storage configs for https://huggingface.co/RedHatAI/Qwen2.5-7B-Instruct-FP8-dynamic
+model: hf
+data: hf

RedHatAI/Voxtral-Mini-3B-2507-FP8-dynamic/accuracy/server.yml

@@ -0,0 +1,7 @@
+enable-chunked-prefill: true
+max-model-len: 9000
+tensor-parallel-size: 1
+trust-remote-code: true
+tokenizer_mode: mistral 
+config_format: mistral 
+load_format: mistral

RedHatAI/whisper-large-v3-turbo-FP8-dynamic/accuracy/server.yml

@@ -0,0 +1,2 @@
+max-model-len: 448
+trust-remote-code: true

deepseek-ai/DeepSeek-R1-Distill-Qwen-32B/accuracy/tasks.yml

@@ -0,0 +1,5 @@
+tasks:
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.8271

deepseek-ai/DeepSeek-R1-Distill-Qwen-32B/storage.yml

@@ -0,0 +1,3 @@
+# storage configs for https://huggingface.co/deepseek-ai/DeepSeek-R1-0528
+model: hf
+data: hf

distil-whisper/distil-large-v3/accuracy/server.yml

@@ -0,0 +1,4 @@
+max-model-len: 448
+tensor-parallel-size: 1
+trust-remote-code: true
+uvicorn-log-level: debug

ibm-granite/granite-4.0-h-small/accuracy/model_card_tasks.yml

@@ -0,0 +1,30 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.7278
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.85
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8370
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.8067
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.7062
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8374

ibm-granite/granite-4.0-h-small/accuracy/tasks.yml

@@ -0,0 +1,60 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.715
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.85
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8573
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.8109
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.6409
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8374
+
+  - name: leaderboard_ifeval
+    metrics:
+      - name: inst_level_strict_acc,none
+        value: 0.729
+
+  # - name: leaderboard_bbh
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.5319
+
+  # - name: leaderboard_math_hard
+  #   metrics:
+  #     - name: exact_match,none
+  #       value: 0.1477
+
+  # - name: leaderboard_gpqa
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.3176
+
+  # - name: leaderboard_musr
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.4601
+
+  - name: leaderboard_mmlu_pro
+    metrics:
+      - name: acc,none
+        value: 0.5545

ibm-granite/granite-4.0-h-small/storage.yml

@@ -0,0 +1,3 @@
+# storage configs for https://huggingface.co/ibm-granite/granite-3.1-8b-instruct
+model: hf
+data: hf

ibm-granite/granite-4.0-micro/accuracy/model_card_tasks.yml

@@ -0,0 +1,30 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.7278
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.79
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8370
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.8067
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.7062
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8374

ibm-granite/granite-4.0-micro/accuracy/tasks.yml

@@ -0,0 +1,60 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.715
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.79
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8573
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.8109
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.6409
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8374
+
+  - name: leaderboard_ifeval
+    metrics:
+      - name: inst_level_strict_acc,none
+        value: 0.729
+
+  # - name: leaderboard_bbh
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.5319
+
+  # - name: leaderboard_math_hard
+  #   metrics:
+  #     - name: exact_match,none
+  #       value: 0.1477
+
+  # - name: leaderboard_gpqa
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.3176
+
+  # - name: leaderboard_musr
+  #   metrics:
+  #     - name: acc-norm,none
+  #       value: 0.4601
+
+  - name: leaderboard_mmlu_pro
+    metrics:
+      - name: acc,none
+        value: 0.5545

ibm-granite/granite-4.0-micro/storage.yml

@@ -0,0 +1,3 @@
+# storage configs for https://huggingface.co/ibm-granite/granite-3.1-8b-instruct
+model: hf
+data: hf

microsoft/Phi-3-medium-4k-instruct/accuracy/model_card_tasks.yml

@@ -0,0 +1,30 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.6442
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.85
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8437
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.803
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.5937
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8058

microsoft/Phi-3-medium-4k-instruct/accuracy/tasks.yml

@@ -0,0 +1,55 @@
+tasks:
+  - name: arc_challenge
+    metrics:
+      - name: acc_norm,none
+        value: 0.6825
+
+  - name: gsm8k
+    metrics:
+      - name: exact_match,strict-match
+        value: 0.82
+
+  - name: hellaswag
+    metrics:
+      - name: acc_norm,none
+        value: 0.8435
+
+  - name: mmlu
+    metrics:
+      - name: acc,none
+        value: 0.803
+
+  - name: truthfulqa_mc2
+    metrics:
+      - name: acc,none
+        value: 0.5934
+
+  - name: winogrande
+    metrics:
+      - name: acc,none
+        value: 0.8011
+
+  - name: leaderboard_ifeval
+    metrics:
+      - name: inst_level_strict_acc,none
+        value: 0.0587
+
+  - name: leaderboard_gpqa_diamond
+    metrics:
+      - name: acc-norm,none
+        value: 0.3939
+
+  - name: leaderboard_gpqa_extended
+    metrics:
+      - name: acc-norm,none
+        value: 0.3882
+
+  - name: leaderboard_gpqa_main
+    metrics:
+      - name: acc-norm,none
+        value: 0.4129
+
+  - name: leaderboard_mmlu_pro
+    metrics:
+      - name: acc,none
+        value: 0.53

mistralai/Voxtral-Mini-3B-2507/accuracy/server.yml

@@ -0,0 +1,7 @@
+enable-chunked-prefill: true
+max-model-len: 9000
+tensor-parallel-size: 1
+trust-remote-code: true
+tokenizer_mode: mistral
+config_format: mistral
+load_format: mistral

mistralai/Voxtral-Small-24B-2507/performance/server.yml

@@ -0,0 +1,8 @@
+enable-chunked-prefill: true
+max-model-len: 9000
+tensor-parallel-size: 1
+trust-remote-code: true
+uvicorn-log-level: debug
+tokenizer_mode: mistral 
+config_format: mistral 
+load_format: mistral

mistralai/Voxtral-Small-24B-2507/storage.yml

@@ -0,0 +1,3 @@
+# https://huggingface.co/mistralai/Voxtral-Small-24B-2507
+model: hf
+data: hf

openai/gpt-oss-120b/performance/server.yml

@@ -0,0 +1,7 @@
+enable-chunked-prefill: true
+max-model-len: 10000
+tensor-parallel-size: 2
+trust-remote-code: true
+uvicorn-log-level: debug
+tool-call-parser: openai 
+enable-auto-tool-choice: true

openai/gpt-oss-20b/accuracy/tasks.yml

@@ -2,4 +2,4 @@ tasks:
   - name: gsm8k
     metrics:
       - name: exact_match,strict-match
-        value: 0
+        value: 0.2494