tessl install tessl/pypi-vllm@0.10.0A high-throughput and memory-efficient inference and serving engine for LLMs
Agent Success
Agent success rate when using this tile
69%
Improvement
Agent success rate improvement when using this tile compared to baseline
1.33x
Baseline
Agent success rate without this tile
52%
{
"context": "This evaluation assesses how well the engineer uses vLLM's SamplingParams class to control text generation behavior. The focus is on proper usage of sampling parameters including temperature, top_p, top_k, penalties, multiple completions (n parameter), seed for reproducibility, and max_tokens for length control.",
"type": "weighted_checklist",
"checklist": [
{
"name": "LLM Initialization",
"description": "Uses vLLM's LLM class to initialize the model, passing the model name/path to the constructor",
"max_score": 10
},
{
"name": "SamplingParams Import",
"description": "Imports and uses the SamplingParams class from vLLM to configure generation parameters",
"max_score": 10
},
{
"name": "Default Generation",
"description": "Implements greedy decoding by setting temperature=0 in SamplingParams for deterministic generation in generate_default function",
"max_score": 15
},
{
"name": "Temperature Parameter",
"description": "Uses the temperature parameter in SamplingParams to control randomness in generate_creative function",
"max_score": 15
},
{
"name": "Top-p Parameter",
"description": "Uses the top_p parameter in SamplingParams for nucleus sampling in generate_creative function",
"max_score": 15
},
{
"name": "Multiple Completions",
"description": "Uses the n parameter in SamplingParams to generate multiple completions per prompt in generate_multiple function",
"max_score": 15
},
{
"name": "Seed Parameter",
"description": "Uses the seed parameter in SamplingParams for reproducible generation in generate_reproducible function",
"max_score": 10
},
{
"name": "Max Tokens",
"description": "Uses the max_tokens parameter in SamplingParams to control output length",
"max_score": 5
},
{
"name": "Generation Method",
"description": "Calls LLM.generate() method with prompts and SamplingParams to produce outputs",
"max_score": 5
}
]
}