From 8563977a107a1b1023beb63cf730d408ec61dd6b Mon Sep 17 00:00:00 2001 From: Zack Zhiyuan Li Date: Sat, 9 Nov 2024 00:42:54 +0000 Subject: [PATCH 1/2] add q8_0 for qwenaudio --- nexa/constants.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/nexa/constants.py b/nexa/constants.py index 32e6a53a..3c7df173 100644 --- a/nexa/constants.py +++ b/nexa/constants.py @@ -112,6 +112,7 @@ class ModelType(Enum): NEXA_RUN_MODEL_MAP_AUDIO_LM = { "qwen2audio": "Qwen2-Audio-7.8B-Instruct:model-q4_0", "qwen2audio:fp16": "Qwen2-Audio-7.8B-Instruct:model-fp16", + "qwen2audio:q8_0": "Qwen2-Audio-7.8B-Instruct:model-q8_0", "qwen2audio:q4_0": "Qwen2-Audio-7.8B-Instruct:model-q4_0", "omniaudio": "Octo-omni-audio:model-fp16", "omniaudio:fp16": "Octo-omni-audio:model-fp16", @@ -185,6 +186,7 @@ class ModelType(Enum): NEXA_RUN_AUDIO_LM_PROJECTOR_MAP = { "qwen2audio": "Qwen2-Audio-7.8B-Instruct:projector-q4_0", "qwen2audio:fp16": "Qwen2-Audio-7.8B-Instruct:projector-fp16", + "qwen2audio:q8_0": "Qwen2-Audio-7.8B-Instruct:projector-q8_0", "qwen2audio:q4_0": "Qwen2-Audio-7.8B-Instruct:projector-q4_0", "omniaudio": "Octo-omni-audio:projector-fp16", "omniaudio:fp16": "Octo-omni-audio:projector-fp16", From 47796f5f5e1072ef0a0330d998273f3d6be21bd8 Mon Sep 17 00:00:00 2001 From: Zack Zhiyuan Li Date: Sat, 9 Nov 2024 00:48:00 +0000 Subject: [PATCH 2/2] make llama.cpp in master-release branch --- dependency/llama.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dependency/llama.cpp b/dependency/llama.cpp index 38c6fa3b..b535cd94 160000 --- a/dependency/llama.cpp +++ b/dependency/llama.cpp @@ -1 +1 @@ -Subproject commit 38c6fa3b8fb6c88075102fd859d04eaea27aa87c +Subproject commit b535cd941e657ac1984d8022dd5f0c98f2b9e265