composte-hermes.yml aktualisiert
This commit is contained in:
parent
3514e0a3e3
commit
9eed8c758c
|
|
@ -26,7 +26,8 @@ services:
|
||||||
- --quantization=awq
|
- --quantization=awq
|
||||||
- --tensor-parallel-size=2
|
- --tensor-parallel-size=2
|
||||||
- --gpu-memory-utilization=0.82
|
- --gpu-memory-utilization=0.82
|
||||||
- --max-model-len=32768
|
- --max-model-len=65536
|
||||||
|
- --rope-scaling={"rope_type":"yarn","factor":2.0,"original_max_position_embeddings":32768}
|
||||||
- --max-num-seqs=16
|
- --max-num-seqs=16
|
||||||
- --kv-cache-dtype=fp8
|
- --kv-cache-dtype=fp8
|
||||||
- --served-model-name=qwen3-14b-awq
|
- --served-model-name=qwen3-14b-awq
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue