MultimodalQnA Image and Audio Support Phase 1 (#1071)

Signed-off-by: Melanie Buehler <[email protected]>
Signed-off-by: okhleif-IL <[email protected]>
Signed-off-by: dmsuehir <[email protected]>
Co-authored-by: Omar Khleif <[email protected]>
Co-authored-by: dmsuehir <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Abolfazl Shahbazi <[email protected]>

Author: 5 people

MultimodalQnA/README.md

@@ -2,7 +2,7 @@
 
 Suppose you possess a set of videos and wish to perform question-answering to extract insights from these videos. To respond to your questions, it typically necessitates comprehension of visual cues within the videos, knowledge derived from the audio content, or often a mix of both these visual elements and auditory facts. The MultimodalQnA framework offers an optimal solution for this purpose.
 
-`MultimodalQnA` addresses your questions by dynamically fetching the most pertinent multimodal information (frames, transcripts, and/or captions) from your collection of videos. For this purpose, MultimodalQnA utilizes [BridgeTower model](https://huggingface.co/BridgeTower/bridgetower-large-itm-mlm-gaudi), a multimodal encoding transformer model which merges visual and textual data into a unified semantic space. During the video ingestion phase, the BridgeTower model embeds both visual cues and auditory facts as texts, and those embeddings are then stored in a vector database. When it comes to answering a question, the MultimodalQnA will fetch its most relevant multimodal content from the vector store and feed it into a downstream Large Vision-Language Model (LVM) as input context to generate a response for the user.
+`MultimodalQnA` addresses your questions by dynamically fetching the most pertinent multimodal information (frames, transcripts, and/or captions) from your collection of videos, images, and audio files. For this purpose, MultimodalQnA utilizes [BridgeTower model](https://huggingface.co/BridgeTower/bridgetower-large-itm-mlm-gaudi), a multimodal encoding transformer model which merges visual and textual data into a unified semantic space. During the ingestion phase, the BridgeTower model embeds both visual cues and auditory facts as texts, and those embeddings are then stored in a vector database. When it comes to answering a question, the MultimodalQnA will fetch its most relevant multimodal content from the vector store and feed it into a downstream Large Vision-Language Model (LVM) as input context to generate a response for the user.
 
 The MultimodalQnA architecture shows below:
 
@@ -100,10 +100,12 @@ In the below, we provide a table that describes for each microservice component
 
 By default, the embedding and LVM models are set to a default value as listed below:
 
-| Service              | Model                                       |
-| -------------------- | ------------------------------------------- |
-| embedding-multimodal | BridgeTower/bridgetower-large-itm-mlm-gaudi |
-| LVM                  | llava-hf/llava-v1.6-vicuna-13b-hf           |
+| Service              | HW    | Model                                     |
+| -------------------- | ----- | ----------------------------------------- |
+| embedding-multimodal | Xeon  | BridgeTower/bridgetower-large-itm-mlm-itc |
+| LVM                  | Xeon  | llava-hf/llava-1.5-7b-hf                  |
+| embedding-multimodal | Gaudi | BridgeTower/bridgetower-large-itm-mlm-itc |
+| LVM                  | Gaudi | llava-hf/llava-v1.6-vicuna-13b-hf         |
 
 You can choose other LVM models, such as `llava-hf/llava-1.5-7b-hf ` and `llava-hf/llava-1.5-13b-hf`, as needed.
 

MultimodalQnA/docker_compose/intel/cpu/xeon/README.md

@@ -84,16 +84,18 @@ export INDEX_NAME="mm-rag-redis"
 export LLAVA_SERVER_PORT=8399
 export LVM_ENDPOINT="http://${host_ip}:8399"
 export EMBEDDING_MODEL_ID="BridgeTower/bridgetower-large-itm-mlm-itc"
+export LVM_MODEL_ID="llava-hf/llava-1.5-7b-hf"
 export WHISPER_MODEL="base"
 export MM_EMBEDDING_SERVICE_HOST_IP=${host_ip}
 export MM_RETRIEVER_SERVICE_HOST_IP=${host_ip}
 export LVM_SERVICE_HOST_IP=${host_ip}
 export MEGA_SERVICE_HOST_IP=${host_ip}
 export BACKEND_SERVICE_ENDPOINT="http://${host_ip}:8888/v1/multimodalqna"
+export DATAPREP_INGEST_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/ingest_with_text"
 export DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_transcripts"
 export DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_captions"
-export DATAPREP_GET_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_videos"
-export DATAPREP_DELETE_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_videos"
+export DATAPREP_GET_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_files"
+export DATAPREP_DELETE_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_files"
 ```
 
 Note: Please replace with `host_ip` with you external IP address, do not use localhost.
@@ -274,54 +276,76 @@ curl http://${host_ip}:9399/v1/lvm \
 
 6. dataprep-multimodal-redis
 
-Download a sample video
+Download a sample video, image, and audio file and create a caption
 
 ```bash
 export video_fn="WeAreGoingOnBullrun.mp4"
 wget http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/WeAreGoingOnBullrun.mp4 -O ${video_fn}
+
+export image_fn="apple.png"
+wget https://github.com/docarray/docarray/blob/main/tests/toydata/image-data/apple.png?raw=true -O ${image_fn}
+
+export caption_fn="apple.txt"
+echo "This is an apple."  > ${caption_fn}
+
+export audio_fn="AudioSample.wav"
+wget https://github.com/intel/intel-extension-for-transformers/raw/main/intel_extension_for_transformers/neural_chat/assets/audio/sample.wav -O ${audio_fn}
 ```
 
-Test dataprep microservice. This command updates a knowledge base by uploading a local video .mp4.
+Test dataprep microservice with generating transcript. This command updates a knowledge base by uploading a local video .mp4 and an audio .wav file.
 
 ```bash
 curl --silent --write-out "HTTPSTATUS:%{http_code}" \
     ${DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT} \
     -H 'Content-Type: multipart/form-data' \
-    -X POST -F "files=@./${video_fn}"
+    -X POST \
+    -F "files=@./${video_fn}" \
+    -F "files=@./${audio_fn}"
 ```
 
-Also, test dataprep microservice with generating caption using lvm microservice
+Also, test dataprep microservice with generating an image caption using lvm microservice
 
 ```bash
 curl --silent --write-out "HTTPSTATUS:%{http_code}" \
     ${DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT} \
     -H 'Content-Type: multipart/form-data' \
-    -X POST -F "files=@./${video_fn}"
+    -X POST -F "files=@./${image_fn}"
+```
+
+Now, test the microservice with posting a custom caption along with an image
+
+```bash
+curl --silent --write-out "HTTPSTATUS:%{http_code}" \
+    ${DATAPREP_INGEST_SERVICE_ENDPOINT} \
+    -H 'Content-Type: multipart/form-data' \
+    -X POST -F "files=@./${image_fn}" -F "files=@./${caption_fn}"
 ```
 
-Also, you are able to get the list of all videos that you uploaded:
+Also, you are able to get the list of all files that you uploaded:
 
 ```bash
 curl -X POST \
     -H "Content-Type: application/json" \
-    ${DATAPREP_GET_VIDEO_ENDPOINT}
+    ${DATAPREP_GET_FILE_ENDPOINT}
 ```
 
-Then you will get the response python-style LIST like this. Notice the name of each uploaded video e.g., `videoname.mp4` will become `videoname_uuid.mp4` where `uuid` is a unique ID for each uploaded video. The same video that are uploaded twice will have different `uuid`.
+Then you will get the response python-style LIST like this. Notice the name of each uploaded file e.g., `videoname.mp4` will become `videoname_uuid.mp4` where `uuid` is a unique ID for each uploaded file. The same files that are uploaded twice will have different `uuid`.
 
 ```bash
 [
     "WeAreGoingOnBullrun_7ac553a1-116c-40a2-9fc5-deccbb89b507.mp4",
-    "WeAreGoingOnBullrun_6d13cf26-8ba2-4026-a3a9-ab2e5eb73a29.mp4"
+    "WeAreGoingOnBullrun_6d13cf26-8ba2-4026-a3a9-ab2e5eb73a29.mp4",
+    "apple_fcade6e6-11a5-44a2-833a-3e534cbe4419.png",
+    "AudioSample_976a85a6-dc3e-43ab-966c-9d81beef780c.wav
 ]
 ```
 
-To delete all uploaded videos along with data indexed with `$INDEX_NAME` in REDIS.
+To delete all uploaded files along with data indexed with `$INDEX_NAME` in REDIS.
 
 ```bash
 curl -X POST \
     -H "Content-Type: application/json" \
-    ${DATAPREP_DELETE_VIDEO_ENDPOINT}
+    ${DATAPREP_DELETE_FILE_ENDPOINT}
 ```
 
 7. MegaService

MultimodalQnA/docker_compose/intel/cpu/xeon/compose.yaml

@@ -36,6 +36,7 @@ services:
       http_proxy: ${http_proxy}
       https_proxy: ${https_proxy}
       PORT: ${EMBEDDER_PORT}
+    entrypoint: ["python", "bridgetower_server.py", "--device", "cpu", "--model_name_or_path", $EMBEDDING_MODEL_ID]
     restart: unless-stopped
   embedding-multimodal:
     image: ${REGISTRY:-opea}/embedding-multimodal:${TAG:-latest}
@@ -76,6 +77,7 @@ services:
       no_proxy: ${no_proxy}
       http_proxy: ${http_proxy}
       https_proxy: ${https_proxy}
+    entrypoint: ["python", "llava_server.py", "--device", "cpu", "--model_name_or_path", $LVM_MODEL_ID]
     restart: unless-stopped
   lvm-llava-svc:
     image: ${REGISTRY:-opea}/lvm-llava-svc:${TAG:-latest}
@@ -125,6 +127,7 @@ services:
       - https_proxy=${https_proxy}
       - http_proxy=${http_proxy}
       - BACKEND_SERVICE_ENDPOINT=${BACKEND_SERVICE_ENDPOINT}
+      - DATAPREP_INGEST_SERVICE_ENDPOINT=${DATAPREP_INGEST_SERVICE_ENDPOINT}
       - DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT=${DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT}
       - DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT=${DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT}
     ipc: host

MultimodalQnA/docker_compose/intel/cpu/xeon/set_env.sh

@@ -15,13 +15,15 @@ export INDEX_NAME="mm-rag-redis"
 export LLAVA_SERVER_PORT=8399
 export LVM_ENDPOINT="http://${host_ip}:8399"
 export EMBEDDING_MODEL_ID="BridgeTower/bridgetower-large-itm-mlm-itc"
+export LVM_MODEL_ID="llava-hf/llava-1.5-7b-hf"
 export WHISPER_MODEL="base"
 export MM_EMBEDDING_SERVICE_HOST_IP=${host_ip}
 export MM_RETRIEVER_SERVICE_HOST_IP=${host_ip}
 export LVM_SERVICE_HOST_IP=${host_ip}
 export MEGA_SERVICE_HOST_IP=${host_ip}
 export BACKEND_SERVICE_ENDPOINT="http://${host_ip}:8888/v1/multimodalqna"
+export DATAPREP_INGEST_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/ingest_with_text"
 export DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_transcripts"
 export DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_captions"
-export DATAPREP_GET_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_videos"
-export DATAPREP_DELETE_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_videos"
+export DATAPREP_GET_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_files"
+export DATAPREP_DELETE_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_files"

MultimodalQnA/docker_compose/intel/hpu/gaudi/README.md

@@ -40,10 +40,11 @@ export MM_RETRIEVER_SERVICE_HOST_IP=${host_ip}
 export LVM_SERVICE_HOST_IP=${host_ip}
 export MEGA_SERVICE_HOST_IP=${host_ip}
 export BACKEND_SERVICE_ENDPOINT="http://${host_ip}:8888/v1/multimodalqna"
+export DATAPREP_INGEST_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/ingest_with_text"
 export DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_transcripts"
 export DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_captions"
-export DATAPREP_GET_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_videos"
-export DATAPREP_DELETE_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_videos"
+export DATAPREP_GET_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_files"
+export DATAPREP_DELETE_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_files"
 ```
 
 Note: Please replace with `host_ip` with you external IP address, do not use localhost.
@@ -224,56 +225,76 @@ curl http://${host_ip}:9399/v1/lvm \
 
 6. Multimodal Dataprep Microservice
 
-Download a sample video
+Download a sample video, image, and audio file and create a caption
 
 ```bash
 export video_fn="WeAreGoingOnBullrun.mp4"
 wget http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/WeAreGoingOnBullrun.mp4 -O ${video_fn}
-```
 
-Test dataprep microservice. This command updates a knowledge base by uploading a local video .mp4.
+export image_fn="apple.png"
+wget https://github.com/docarray/docarray/blob/main/tests/toydata/image-data/apple.png?raw=true -O ${image_fn}
+
+export caption_fn="apple.txt"
+echo "This is an apple."  > ${caption_fn}
+
+export audio_fn="AudioSample.wav"
+wget https://github.com/intel/intel-extension-for-transformers/raw/main/intel_extension_for_transformers/neural_chat/assets/audio/sample.wav -O ${audio_fn}
+```
 
-Test dataprep microservice with generating transcript using whisper model
+Test dataprep microservice with generating transcript. This command updates a knowledge base by uploading a local video .mp4 and an audio .wav file.
 
 ```bash
 curl --silent --write-out "HTTPSTATUS:%{http_code}" \
     ${DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT} \
     -H 'Content-Type: multipart/form-data' \
-    -X POST -F "files=@./${video_fn}"
+    -X POST \
+    -F "files=@./${video_fn}" \
+    -F "files=@./${audio_fn}"
 ```
 
-Also, test dataprep microservice with generating caption using lvm-tgi
+Also, test dataprep microservice with generating an image caption using lvm-tgi
 
 ```bash
 curl --silent --write-out "HTTPSTATUS:%{http_code}" \
     ${DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT} \
     -H 'Content-Type: multipart/form-data' \
-    -X POST -F "files=@./${video_fn}"
+    -X POST -F "files=@./${image_fn}"
+```
+
+Now, test the microservice with posting a custom caption along with an image
+
+```bash
+curl --silent --write-out "HTTPSTATUS:%{http_code}" \
+    ${DATAPREP_INGEST_SERVICE_ENDPOINT} \
+    -H 'Content-Type: multipart/form-data' \
+    -X POST -F "files=@./${image_fn}" -F "files=@./${caption_fn}"
 ```
 
-Also, you are able to get the list of all videos that you uploaded:
+Also, you are able to get the list of all files that you uploaded:
 
 ```bash
 curl -X POST \
     -H "Content-Type: application/json" \
-    ${DATAPREP_GET_VIDEO_ENDPOINT}
+    ${DATAPREP_GET_FILE_ENDPOINT}
 ```
 
-Then you will get the response python-style LIST like this. Notice the name of each uploaded video e.g., `videoname.mp4` will become `videoname_uuid.mp4` where `uuid` is a unique ID for each uploaded video. The same video that are uploaded twice will have different `uuid`.
+Then you will get the response python-style LIST like this. Notice the name of each uploaded file e.g., `videoname.mp4` will become `videoname_uuid.mp4` where `uuid` is a unique ID for each uploaded file. The same files that are uploaded twice will have different `uuid`.
 
 ```bash
 [
     "WeAreGoingOnBullrun_7ac553a1-116c-40a2-9fc5-deccbb89b507.mp4",
-    "WeAreGoingOnBullrun_6d13cf26-8ba2-4026-a3a9-ab2e5eb73a29.mp4"
+    "WeAreGoingOnBullrun_6d13cf26-8ba2-4026-a3a9-ab2e5eb73a29.mp4",
+    "apple_fcade6e6-11a5-44a2-833a-3e534cbe4419.png",
+    "AudioSample_976a85a6-dc3e-43ab-966c-9d81beef780c.wav
 ]
 ```
 
-To delete all uploaded videos along with data indexed with `$INDEX_NAME` in REDIS.
+To delete all uploaded files along with data indexed with `$INDEX_NAME` in REDIS.
 
 ```bash
 curl -X POST \
     -H "Content-Type: application/json" \
-    ${DATAPREP_DELETE_VIDEO_ENDPOINT}
+    ${DATAPREP_DELETE_FILE_ENDPOINT}
 ```
 
 7. MegaService

MultimodalQnA/docker_compose/intel/hpu/gaudi/compose.yaml

@@ -36,6 +36,7 @@ services:
       http_proxy: ${http_proxy}
       https_proxy: ${https_proxy}
       PORT: ${EMBEDDER_PORT}
+    entrypoint: ["python", "bridgetower_server.py", "--device", "hpu", "--model_name_or_path", $EMBEDDING_MODEL_ID]
     restart: unless-stopped
   embedding-multimodal:
     image: ${REGISTRY:-opea}/embedding-multimodal:${TAG:-latest}
@@ -139,6 +140,7 @@ services:
       - https_proxy=${https_proxy}
       - http_proxy=${http_proxy}
       - BACKEND_SERVICE_ENDPOINT=${BACKEND_SERVICE_ENDPOINT}
+      - DATAPREP_INGEST_SERVICE_ENDPOINT=${DATAPREP_INGEST_SERVICE_ENDPOINT}
       - DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT=${DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT}
       - DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT=${DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT}
     ipc: host

MultimodalQnA/docker_compose/intel/hpu/gaudi/set_env.sh

@@ -22,7 +22,8 @@ export MM_RETRIEVER_SERVICE_HOST_IP=${host_ip}
 export LVM_SERVICE_HOST_IP=${host_ip}
 export MEGA_SERVICE_HOST_IP=${host_ip}
 export BACKEND_SERVICE_ENDPOINT="http://${host_ip}:8888/v1/multimodalqna"
+export DATAPREP_INGEST_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/ingest_with_text"
 export DATAPREP_GEN_TRANSCRIPT_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_transcripts"
 export DATAPREP_GEN_CAPTION_SERVICE_ENDPOINT="http://${host_ip}:6007/v1/generate_captions"
-export DATAPREP_GET_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_videos"
-export DATAPREP_DELETE_VIDEO_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_videos"
+export DATAPREP_GET_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/get_files"
+export DATAPREP_DELETE_FILE_ENDPOINT="http://${host_ip}:6007/v1/dataprep/delete_files"