results_arcee_supernova_medius.py

"""Module containing performance results for the Arcee-SuperNova-Medius model."""

results_arcee_supernova_medius = {
    "name": "Arcee-SuperNova-Medius",
    "modelType": "Qwen2.5 14B",
    "configurations": [
        {
            "instanceType": "r8g.4xlarge",
            "quantization": "Q8_0",
            "container": "llama.cpp 11/27/24",
            "status": "OK",
            "tokensPerSecond": "13.5",
            "notes": "-fa",
        },
        {
            "instanceType": "r8g.4xlarge",
            "quantization": "Q4_0_4_8",
            "container": "llama.cpp 11/27/24",
            "status": "OK",
            "tokensPerSecond": "22",
            "notes": "-fa",
        },
        {
            "instanceType": "g5.12xlarge",
            "quantization": "None",
            "container": "LMI 0.30+vLLM 0.6.2",
            "status": "OK",
            "tokensPerSecond": "45",
            "notes": "",
        },
        {
            "instanceType": "g6.12xlarge",
            "quantization": "None",
            "container": "LMI 0.30+vLLM 0.6.2",
            "status": "OK",
            "tokensPerSecond": "29",
            "notes": "",
        },
        {
            "instanceType": "g6e.12xlarge",
            "quantization": "None",
            "container": "vLLM0.6.4.post1",
            "status": "OK",
            "tokensPerSecond": "70",
            "notes": "--tensor-parallel-size 4 --max-model-len 16384",
        },
        {
            "instanceType": "g6e.12xlarge",
            "quantization": "None",
            "container": "vLLM 0.6.4.post1",
            "status": "OK",
            "tokensPerSecond": "43",
            "notes": "--tensor-parallel-size 2 --max-model-len 16384",
        },
        {
            "instanceType": "p4d.24xlarge",
            "quantization": "None",
            "container": "LMI 0.30+vLLM 0.6.2",
            "status": "OK",
            "tokensPerSecond": "108",
            "notes": "",
        },
        {
            "instanceType": "p5.48xlarge",
            "quantization": "None",
            "container": "LMI 0.30+vLLM 0.6.2",
            "status": "N/A",
            "tokensPerSecond": "N/A",
            "notes": '"OPTION_MAX_MODEL_LEN": "32768",\n"TENSOR_PARALLEL_DEGREE": "max",',
        },
    ],
}