results_arcee_lite.py

"""Module containing performance results for the Arcee-Lite model."""

results_arcee_lite = {
    "name": "Arcee-Lite",
    "modelType": "Qwen2 1.5B distilled from phi-3-medium 14B",
    "configurations": [
        {
            "instanceType": "c6i.xlarge",
            "quantization": "bitsandbytes-nf4",
            "container": "TGI 2.2.0",
            "status": "KO",
            "tokensPerSecond": "-",
            "notes": "OOM, might work with a prequantized model",
        },
        {
            "instanceType": "c6i.2xlarge",
            "quantization": "bitsandbytes-nf4",
            "container": "TGI 2.2.0",
            "status": "KO",
            "tokensPerSecond": "-",
            "notes": "OOM, might work with a prequantized model",
        },
        {
            "instanceType": "c6i.4xlarge",
            "configurations": [
                {
                    "quantization": "none",
                    "container": "TGI 2.2.0",
                    "status": "OK",
                    "tokensPerSecond": "10.7",
                },
                {
                    "quantization": "bitsandbytes (int8)",
                    "container": "TGI 2.2.0",
                    "status": "OK",
                    "tokensPerSecond": "10.5",
                },
                {
                    "quantization": "bitsandbytes-nf4",
                    "container": "TGI 2.2.0",
                    "status": "OK",
                    "tokensPerSecond": "10.6",
                },
            ],
        },
        {
            "instanceType": "c7i.4xlarge",
            "quantization": "none",
            "container": "TGI 2.2.0",
            "status": "waiting for quota",
            "tokensPerSecond": "-",
        },
        {
            "instanceType": "g5.xlarge",
            "configurations": [
                {
                    "quantization": "none",
                    "container": "TGI 2.2.0",
                    "status": "OK",
                    "tokensPerSecond": "110",
                },
                {
                    "quantization": "none",
                    "container": "DJL 0.28 vLLM",
                    "status": "OK",
                    "tokensPerSecond": "105",
                    "notes": '"OPTION_MAX_MODEL_LEN": "32768",',
                },
            ],
        },
        {
            "instanceType": "g6e.2xlarge",
            "configurations": [
                {
                    "container": "TGI 2.2.0",
                    "quantization": "none",
                    "status": "OK",
                    "tokensPerSecond": "160",
                },
                {
                    "container": "SGLang 0.2.13",
                    "quantization": "none",
                    "status": "OK",
                    "tokensPerSecond": "167",
                },
                {
                    "container": "vLLM 0.5.5",
                    "quantization": "none",
                    "status": "OK",
                    "tokensPerSecond": "150",
                },
            ],
        },
        {
            "instanceType": "c7g.2xlarge",
            "configurations": [
                {
                    "container": "PyTorch 2.4.1 on EC2",
                    "quantization": "none",
                    "status": "OK",
                    "tokensPerSecond": "16.5",
                },
            ],
        },
        {
            "instanceType": "c7g.8xlarge",
            "configurations": [
                {
                    "container": "PyTorch 2.4.1 on EC2",
                    "quantization": "none",
                    "status": "OK",
                    "tokensPerSecond": "26",
                },
            ],
        },
    ],
}