fix(autogptq): do not use_triton with qwen-vl (#1985)

* Enhance autogptq backend to support VL models

* update dependencies for autogptq

* remove redundant auto-gptq dependency

* Convert base64 to image_url for Qwen-VL model

* implemented model inference for qwen-vl

* remove user prompt from generated answer

* fixed write image error

* fixed use_triton issue when loading Qwen-VL model

---------

Co-authored-by: Binghua Wu <bingwu@estee.com>

.github/workflows/image_build.yml

@@ -280,6 +280,7 @@ jobs:
         run: |
           docker pull localai/localai:${{ steps.meta.outputs.version }}
           docker tag localai/localai:${{ steps.meta.outputs.version }} localai/localai:${{ inputs.latest-image }}
+          docker push localai/localai:${{ inputs.latest-image }}
           docker pull quay.io/go-skynet/local-ai:${{ steps.meta.outputs.version }}
           docker tag quay.io/go-skynet/local-ai:${{ steps.meta.outputs.version }} quay.io/go-skynet/local-ai:${{ inputs.latest-image }}
           docker push quay.io/go-skynet/local-ai:${{ inputs.latest-image }}
@@ -289,6 +290,7 @@ jobs:
         run: |
           docker pull localai/localai:${{ steps.meta_aio_dockerhub.outputs.version }}
           docker tag localai/localai:${{ steps.meta_aio_dockerhub.outputs.version }} localai/localai:${{ inputs.latest-image-aio }}
+          docker push localai/localai:${{ inputs.latest-image-aio }}
           docker pull quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }}
           docker tag quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }} quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}
           docker push quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}

Makefile

@@ -5,7 +5,7 @@ BINARY_NAME=local-ai
 
 # llama.cpp versions
 GOLLAMA_STABLE_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
-CPPLLAMA_VERSION?=cc4a95426d17417d3c83f12bdb514fbe8abe2a88
+CPPLLAMA_VERSION?=1b67731e184e27a465b8c5476061294a4af668ea
 
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
@@ -16,7 +16,7 @@ RWKV_REPO?=https://github.com/donomii/go-rwkv.cpp
 RWKV_VERSION?=661e7ae26d442f5cfebd2a0881b44e8c55949ec6
 
 # whisper.cpp version
-WHISPER_CPP_VERSION?=13c22321d1ac758ce68a429c23104e234b440769
+WHISPER_CPP_VERSION?=8f253ef3af1c62c04316ba4afa7145fc4d701a8c
 
 # bert.cpp version
 BERT_VERSION?=6abe312cded14042f6b7c3cd8edf082713334a4d

backend/python/autogptq/autogptq.py

@@ -39,7 +39,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                 self.model_name = "Qwen-VL-Chat"
                 model = AutoModelForCausalLM.from_pretrained(model_path, 
                     trust_remote_code=request.TrustRemoteCode,
-                    use_triton=request.UseTriton,
                     device_map="auto").eval()
             else:
                 model = AutoGPTQForCausalLM.from_quantized(model_path,

docs/data/version.json

@@ -1,3 +1,3 @@
 {
-  "version": "v2.11.0"
+  "version": "v2.12.1"
 }