diff --git a/Makefile b/Makefile index dfa91a15..7aaad492 100644 --- a/Makefile +++ b/Makefile @@ -8,7 +8,7 @@ DETECT_LIBS?=true # llama.cpp versions GOLLAMA_REPO?=https://github.com/go-skynet/go-llama.cpp GOLLAMA_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be -CPPLLAMA_VERSION?=a1649cc13f89946322358f92ea268ae1b7b5096c +CPPLLAMA_VERSION?=92bc493917d43b83e592349e138b54c90b1c3ea7 # whisper.cpp version WHISPER_REPO?=https://github.com/ggerganov/whisper.cpp diff --git a/gallery/index.yaml b/gallery/index.yaml index 3a9d7abe..fb5476f9 100644 --- a/gallery/index.yaml +++ b/gallery/index.yaml @@ -5589,6 +5589,33 @@ - filename: marco-o1-uncensored.Q4_K_M.gguf sha256: ad0440270a7254098f90779744d3e5b34fe49b7baf97c819909ba9c5648cc0d9 uri: huggingface://QuantFactory/marco-o1-uncensored-GGUF/marco-o1-uncensored.Q4_K_M.gguf +- !!merge <<: *qwen2 + name: "minicpm-v-2_6" + license: apache-2.0 + icon: https://avatars.githubusercontent.com/u/89920203 + urls: + - https://huggingface.co/openbmb/MiniCPM-V-2_6-gguf + - https://huggingface.co/openbmb/MiniCPM-V-2_6 + description: | + MiniCPM-V 2.6 is the latest and most capable model in the MiniCPM-V series. The model is built on SigLip-400M and Qwen2-7B with a total of 8B parameters + tags: + - llm + - multimodal + - gguf + - gpu + - qwen2 + - cpu + overrides: + mmproj: minicpm-v-2_6-mmproj-f16.gguf + parameters: + model: minicpm-v-2_6-Q4_K_M.gguf + files: + - filename: minicpm-v-2_6-Q4_K_M.gguf + sha256: 3a4078d53b46f22989adbf998ce5a3fd090b6541f112d7e936eb4204a04100b1 + uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/ggml-model-Q4_K_M.gguf + - filename: minicpm-v-2_6-mmproj-f16.gguf + sha256: f8a805e9e62085805c69c427287acefc284932eb4abfe6e1b1ce431d27e2f4e0 + uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/mmproj-model-f16.gguf - &mistral03 ## START Mistral url: "github:mudler/LocalAI/gallery/mistral-0.3.yaml@master" @@ -9145,7 +9172,7 @@ urls: - https://huggingface.co/BAAI/Bunny-Llama-3-8B-V-gguf description: | - Bunny is a family of lightweight but powerful multimodal models. It offers multiple plug-and-play vision encoders, like EVA-CLIP, SigLIP and language backbones, including Llama-3-8B, Phi-1.5, StableLM-2, Qwen1.5, MiniCPM and Phi-2. To compensate for the decrease in model size, we construct more informative training data by curated selection from a broader data source. + Bunny is a family of lightweight but powerful multimodal models. It offers multiple plug-and-play vision encoders, like EVA-CLIP, SigLIP and language backbones, including Llama-3-8B, Phi-1.5, StableLM-2, Qwen1.5, and Phi-2. To compensate for the decrease in model size, we construct more informative training data by curated selection from a broader data source. We provide Bunny-Llama-3-8B-V, which is built upon SigLIP and Llama-3-8B-Instruct. More details about this model can be found in GitHub. icon: https://huggingface.co/BAAI/Bunny-Llama-3-8B-V-gguf/resolve/main/icon.png @@ -9218,33 +9245,6 @@ - filename: minicpm-llama3-mmproj-f16.gguf sha256: 391d11736c3cd24a90417c47b0c88975e86918fcddb1b00494c4d715b08af13e uri: huggingface://openbmb/MiniCPM-Llama3-V-2_5-gguf/mmproj-model-f16.gguf -- !!merge <<: *llama3 - name: "minicpm-v-2_6" - license: apache-2.0 - icon: https://avatars.githubusercontent.com/u/89920203 - urls: - - https://huggingface.co/openbmb/MiniCPM-V-2_6-gguf - - https://huggingface.co/openbmb/MiniCPM-V-2_6 - description: | - MiniCPM-V 2.6 is the latest and most capable model in the MiniCPM-V series. The model is built on SigLip-400M and Qwen2-7B with a total of 8B parameters - tags: - - llm - - multimodal - - gguf - - gpu - - llama3 - - cpu - overrides: - mmproj: minicpm-v-2_6-mmproj-f16.gguf - parameters: - model: minicpm-v-2_6-Q4_K_M.gguf - files: - - filename: minicpm-v-2_6-Q4_K_M.gguf - sha256: 3a4078d53b46f22989adbf998ce5a3fd090b6541f112d7e936eb4204a04100b1 - uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/ggml-model-Q4_K_M.gguf - - filename: minicpm-v-2_6-mmproj-f16.gguf - sha256: f8a805e9e62085805c69c427287acefc284932eb4abfe6e1b1ce431d27e2f4e0 - uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/mmproj-model-f16.gguf - !!merge <<: *llama3 name: "llama-3-cursedstock-v1.8-8b-iq-imatrix" urls: @@ -10108,7 +10108,7 @@ urls: - https://huggingface.co/internlm/internlm2_5-7b-chat-1m - https://huggingface.co/bartowski/internlm2_5-7b-chat-1m-GGUF - icon: https://github.com/InternLM/InternLM/assets/22529082/b9788105-8892-4398-8b47-b513a292378e + icon: https://avatars.githubusercontent.com/u/135356492 tags: - internlm2 - gguf @@ -10129,6 +10129,31 @@ - filename: internlm2_5-7b-chat-1m-Q4_K_M.gguf uri: huggingface://bartowski/internlm2_5-7b-chat-1m-GGUF/internlm2_5-7b-chat-1m-Q4_K_M.gguf sha256: 10d5e18a4125f9d4d74a9284a21e0c820b150af06dee48665e54ff6e1be3a564 +### Internlm3 +- name: "internlm3-8b-instruct" + url: "github:mudler/LocalAI/gallery/chatml.yaml@master" + urls: + - https://huggingface.co/internlm/internlm3-8b-instruct + - https://huggingface.co/bartowski/internlm3-8b-instruct-GGUF + icon: https://avatars.githubusercontent.com/u/135356492 + tags: + - internlm3 + - gguf + - cpu + - gpu + description: | + InternLM3 has open-sourced an 8-billion parameter instruction model, InternLM3-8B-Instruct, designed for general-purpose usage and advanced reasoning. The model has the following characteristics: + + Enhanced performance at reduced cost: State-of-the-art performance on reasoning and knowledge-intensive tasks surpass models like Llama3.1-8B and Qwen2.5-7B. + + Deep thinking capability: InternLM3 supports both the deep thinking mode for solving complicated reasoning tasks via the long chain-of-thought and the normal response mode for fluent user interactions. + overrides: + parameters: + model: internlm3-8b-instruct-Q4_K_M.gguf + files: + - filename: internlm3-8b-instruct-Q4_K_M.gguf + uri: huggingface://bartowski/internlm3-8b-instruct-GGUF/internlm3-8b-instruct-Q4_K_M.gguf + sha256: 2a9644687318e8659c9cf9b40730d5cc2f5af06f786a50439c7c51359b23896e - &phi-3 ### START Phi-3 url: "github:mudler/LocalAI/gallery/phi-3-chat.yaml@master"