From 1e98eea3bd004eee888ad668c1551783b59df699 Mon Sep 17 00:00:00 2001 From: wmantly Date: Tue, 22 Sep 2026 01:22:46 +0000 Subject: [PATCH] chore(config): switch model to Qwen3.8-27B Q4_K_P with FastMTP 32K draft --- scripts/ollama-proxy.py | 6 +++--- scripts/start-server.sh | 2 +- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/scripts/ollama-proxy.py b/scripts/ollama-proxy.py index 9a43ac6..bfb4206 100644 --- a/scripts/ollama-proxy.py +++ b/scripts/ollama-proxy.py @@ -353,7 +353,7 @@ def _model_entry(tag): "family": "qwen35", "families": ["qwen35", "qwen2", "clip"], "parameter_size": "27.0B", - "quantization_level": "Q5_K_P", + "quantization_level": "Q4_K_P", "context_length": ctx, }, "capabilities": ["completion", "chat", "tools", "thinking", "vision", "multimodal"], @@ -376,7 +376,7 @@ def ollama_ps(): "family": "qwen35", "families": ["qwen35", "qwen2", "clip"], "parameter_size": "27.0B", - "quantization_level": "Q5_K_P", + "quantization_level": "Q4_K_P", "context_length": ctx, }, "expires_at": time.strftime("%Y-%m-%dT%H:%M:%S", time.gmtime(time.time() + 86400)) + ".000000Z", @@ -401,7 +401,7 @@ def ollama_show(payload=None): "family": "qwen35", "families": ["qwen35", "qwen2", "clip"], "parameter_size": "27.0B", - "quantization_level": "Q5_K_P", + "quantization_level": "Q4_K_P", "context_length": ctx, }, "model_info": { diff --git a/scripts/start-server.sh b/scripts/start-server.sh index d221bd3..b882c9b 100755 --- a/scripts/start-server.sh +++ b/scripts/start-server.sh @@ -17,7 +17,7 @@ RTX_DEV=$(/opt/llama.cpp-xrip/build-nccl/bin/llama-cli --list-devices | grep -i RTX_DEV=${RTX_DEV:-CUDA0} exec /opt/llama.cpp-xrip/build-nccl/bin/llama-server \ - -m /opt/models/gguf/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-Q5_K_P.gguf \ + -m /opt/models/gguf/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-Q4_K_P.gguf \ --spec-draft-model /opt/models/gguf/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-FastMTP-32K.gguf \ --spec-draft-device "$RTX_DEV" \ --spec-draft-ngl all \