feat: switch models — ollama qwen3.5:4b-fast, vllm numind/NuExtract3
- Ollama: qwen3.5:4b-fast with NUM_PARALLEL=8, num_ctx=16384 (~6GB VRAM) - vLLM: numind/NuExtract3 at 192.168.42.254:2701, gpu-memory-util=0.45 - All 3 pipelines (beta/paper/live) updated with correct endpoints - Ollama proxy at 10.1.1.12:2701 (K8s ollama-metrics) - vLLM proxy at 192.168.42.254:2701 (NixOS nginx)
This commit is contained in:
@@ -40,9 +40,9 @@ config:
|
||||
BROKER_MODE: "paper"
|
||||
BROKER_PROVIDER: "alpaca"
|
||||
OLLAMA_BASE_URL: "http://10.1.1.12:2701"
|
||||
OLLAMA_MODEL: "qwen3.5:9b-fast"
|
||||
VLLM_BASE_URL: "http://10.1.1.12:2701"
|
||||
VLLM_MODEL: "AxionML/Qwen3.5-9B-NVFP4"
|
||||
OLLAMA_MODEL: "qwen3.5:4b-fast"
|
||||
VLLM_BASE_URL: "http://192.168.42.254:2701"
|
||||
VLLM_MODEL: "numind/NuExtract3"
|
||||
VLLM_TIMEOUT: "120"
|
||||
VLLM_MAX_RETRIES: "2"
|
||||
VLLM_TEMPERATURE: "0.7"
|
||||
|
||||
@@ -17,6 +17,9 @@ config:
|
||||
DEPLOY_STAGE: "paper"
|
||||
POSTGRES_USER: "stonks_paper"
|
||||
OLLAMA_BASE_URL: "http://10.1.1.12:2701"
|
||||
OLLAMA_MODEL: "qwen3.5:4b-fast"
|
||||
VLLM_BASE_URL: "http://192.168.42.254:2701"
|
||||
VLLM_MODEL: "numind/NuExtract3"
|
||||
MARKET_DATA_BASE_URL: "https://api.polygon.io"
|
||||
|
||||
## Secrets — all credentials for the paper stage
|
||||
|
||||
@@ -186,14 +186,14 @@ config:
|
||||
MINIO_ENDPOINT: "minio.minio-service.svc.cluster.local:80"
|
||||
MINIO_SECURE: "false"
|
||||
OLLAMA_BASE_URL: "http://10.1.1.12:2701"
|
||||
OLLAMA_MODEL: "qwen3.5:9b-fast"
|
||||
OLLAMA_MODEL: "qwen3.5:4b-fast"
|
||||
OLLAMA_TIMEOUT: "240"
|
||||
OLLAMA_MAX_RETRIES: "2"
|
||||
OLLAMA_RETRY_BASE_DELAY: "1.0"
|
||||
OLLAMA_RETRY_MAX_DELAY: "10.0"
|
||||
OLLAMA_RETRY_BACKOFF_MULTIPLIER: "2.0"
|
||||
VLLM_BASE_URL: "http://10.1.1.12:2701"
|
||||
VLLM_MODEL: "qwen3.5:9b-fast"
|
||||
VLLM_BASE_URL: "http://192.168.42.254:2701"
|
||||
VLLM_MODEL: "numind/NuExtract3"
|
||||
VLLM_TIMEOUT: "120"
|
||||
VLLM_MAX_RETRIES: "2"
|
||||
VLLM_TEMPERATURE: "0.7"
|
||||
|
||||
Reference in New Issue
Block a user