adding and ensuring consistency of whisper-small and mistral-code models

Signed-off-by: greg pereira <[email protected]>

.github/workflows/codegen.yaml

@@ -63,7 +63,7 @@ jobs:
 
       - name: Download model
         working-directory: ./models
-        run: make download-model-mistral
+        run: make download-model-mistral-code
 
       - name: Run Functional Tests
         shell: bash

.github/workflows/model_image_build_push.yaml

@@ -43,6 +43,9 @@ jobs:
             label: Q4_K_M
             url: https://huggingface.co/instructlab/granite-7b-lab-GGUF/resolve/main/granite-7b-lab-Q4_K_M.gguf
             platforms: linux/amd64,linux/arm64
+          - image_name: whisper-small
+            url: https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin
+            platforms: linux/amd64,linux/arm64
     runs-on: ubuntu-latest
     permissions:
       contents: read

models/Containerfile

@@ -1,8 +1,8 @@
 # Suggested alternative open AI Models
-# 	    https://huggingface.co/instructlab/granite-7b-lab-GGUF/resolve/main/granite-7b-lab-Q4_K_M.gguf
+# 	    https://huggingface.co/instructlab/granite-7b-lab-GGUF/resolve/main/granite-7b-lab-Q4_K_M.gguf (Default)
 #	    https://huggingface.co/instructlab/merlinite-7b-lab-GGUF/resolve/main/merlinite-7b-lab-Q4_K_M.gguf
 # 	    https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/resolve/main/llama-2-7b-chat.Q5_K_S.gguf
-#	    https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_K_M.gguf (Default)
+#	    https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_K_M.gguf
 #	    https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q4_K_M.gguf
 #	    https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin
 # podman build --build-arg="MODEL_URL=https://..." -t quay.io/yourimage .

models/Makefile

@@ -25,15 +25,19 @@ download-model-granite:
 download-model-merlinite:
 	$(MAKE) MODEL_URL=https://huggingface.co/instructlab/merlinite-7b-lab-GGUF/resolve/main/merlinite-7b-lab-Q4_K_M.gguf MODEL_NAME=merlinite-7b-lab-Q4_K_M.gguf download-model
 
-.PHONY: download-model-whisper-small # small .bin model type testing
+.PHONY: download-model-whisper-small
 download-model-whisper-small:
 	$(MAKE) MODEL_NAME=ggml-small.bin MODEL_URL=https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin download-model
 
 .PHONY: download-model-mistral
 download-model-mistral:
 	$(MAKE) MODEL_NAME=mistral-7b-instruct-v0.2.Q4_K_M.gguf MODEL_URL=https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_K_M.gguf download-model
 
+.PHONY: download-model-mistral-code
+download-mode-mistral-code:
+	$(MAKE) MODEL_NAME=mistral-7b-code-16k-qlora.Q4_K_M.gguf MODEL_URL=https://huggingface.co/TheBloke/Mistral-7B-Code-16K-qlora-GGUF/resolve/main/mistral-7b-code-16k-qlora.Q4_K_M.gguf
+
 .PHONY: clean
 clean:
-	rm -f *tmp
-	rm -f mistral* whisper* granite* merlinite*
+	-rm -f *tmp
+	-rm -f mistral* ggml-* granite* merlinite*

models/README.md

@@ -8,6 +8,7 @@ Want to try one of our tested models? Try one or all of the following:
 make download-model-granite
 make download-model-merlinite
 make download-model-mistral
+make download-model-mistral-code
 make download-model-whisper-small
 ```
 

recipes/audio/audio_to_text/bootc/Containerfile

@@ -14,9 +14,9 @@ RUN set -eu; mkdir -p /usr/ssh && \
     echo ${SSHPUBKEY} > /usr/ssh/root.keys && chmod 0600 /usr/ssh/root.keys
 
 ARG RECIPE=audio-to-text
-ARG MODEL_IMAGE=quay.io/ai-lab/mistral-7b-instruct:latest
+ARG MODEL_IMAGE=quay.io/ai-lab/whisper-small:latest
 ARG APP_IMAGE=quay.io/ai-lab/${RECIPE}:latest
-ARG SERVER_IMAGE=quay.io/ai-lab/llamacpp_python:latest
+ARG SERVER_IMAGE=quay.io/ai-lab/whispercpp:latest
 ARG TARGETARCH
 
 # Add quadlet files to setup system to automatically run AI application on boot