MCPcopy
hub / github.com/InternLM/lmdeploy / get_max_cache_entry

Function get_max_cache_entry

autotest/utils/benchmark_utils.py:274–284  ·  view source on GitHub ↗
(model, backend)

Source from the content-addressed store, hash-verified

272
273
274def get_max_cache_entry(model, backend):
275 if backend == 'pytorch':
276 return 0.8
277 if 'Llama-2' in model:
278 return 0.95
279 elif 'internlm2' in model:
280 return 0.9
281 elif 'Qwen/Qwen3-235B-A22B' == model or 'internlm/Intern-S1' == model:
282 return 0.7
283 else:
284 return None

Callers 4

throughput_testFunction · 0.85
longtext_throughput_testFunction · 0.85
restful_testFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected