models(gallery): add Bunny-llama (mudler#2311)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Signed-off-by: Jose Luis López López <joseluisll@gmail.com>

gallery/index.yaml

@@ -852,6 +852,33 @@
     - filename: Llava_1.5_Llama3_mmproj_updated.gguf
       sha256: 4f2bb77ca60f2c932d1c6647d334f5d2cd71966c19e850081030c9883ef1906c
       uri: https://huggingface.co/ChaoticNeutrals/LLaVA-Llama-3-8B-mmproj-Updated/resolve/main/llava-v1.5-8B-Updated-Stop-Token/mmproj-model-f16.gguf
+- !!merge <<: *llama3
+  name: "bunny-llama-3-8b-v"
+  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.
+
+      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
+  tags:
+    - llm
+    - multimodal
+    - gguf
+    - gpu
+    - llama3
+    - cpu
+  overrides:
+    mmproj: Bunny-Llama-3-8B-Q4_K_M-mmproj.gguf
+    parameters:
+      model: Bunny-Llama-3-8B-Q4_K_M.gguf
+  files:
+    - filename: Bunny-Llama-3-8B-Q4_K_M-mmproj.gguf
+      sha256: 96d033387a91e56cf97fa5d60e02c0128ce07c8fa83aaaefb74ec40541615ea5
+      uri: huggingace://BAAI/Bunny-Llama-3-8B-V-gguf/mmproj-model-f16.gguf
+    - filename: Bunny-Llama-3-8B-Q4_K_M.gguf
+      sha256: 88f0a61f947dbf129943328be7262ae82e3a582a0c75e53544b07f70355a7c30
+      uri: huggingace://BAAI/Bunny-Llama-3-8B-V-gguf/ggml-model-Q4_K_M.gguf
 - !!merge <<: *llama3
   name: "llava-llama-3-8b-v1_1"
   description: |