|
|
@@ -12,7 +12,6 @@ engine = "vllm"
|
|
|
name = "DeepSeek V4 Flash DSpark (vLLM)"
|
|
|
docker_image = "aidendle94/sparkrun-vllm-ds4-gb10:production-3.6"
|
|
|
model = "deepseek-ai/DeepSeek-V4-Flash-DSpark"
|
|
|
-port = 8000
|
|
|
required_nodes = 2
|
|
|
tensor_parallel_size = 2
|
|
|
pipeline_parallel_size = 1
|
|
|
@@ -29,7 +28,6 @@ args = [
|
|
|
"--max-num-batched-tokens", "4096",
|
|
|
"--max-cudagraph-capture-size", "128",
|
|
|
"--gpu-memory-utilization", "0.8",
|
|
|
- "--served-model-name", "ChatGPTN",
|
|
|
# Engine backends
|
|
|
"--attention-backend", "FLASHINFER_MLA_SPARSE_DSV4",
|
|
|
"--moe-backend", "b12x",
|
