From d307f26492fcfadebfca88b353007a07c313cd51 Mon Sep 17 00:00:00 2001
From: felix0496 <felix.kirsch@kern.ai>
Date: Thu, 24 Nov 2022 09:53:18 +0100
Subject: [PATCH 1/4] save embedder as pickle

---
 controller.py |  9 +++++++++
 start         | 15 +++++++++++++++
 2 files changed, 24 insertions(+)

diff --git a/controller.py b/controller.py
index 3e5c9e1..b61834e 100644
--- a/controller.py
+++ b/controller.py
@@ -10,6 +10,7 @@
     notification,
     organization,
 )
+import pickle
 import torch
 import traceback
 from requests.exceptions import HTTPError
@@ -449,6 +450,14 @@ def run_encoding(
             request_util.post_embedding_to_neural_search(
                 request.project_id, embedding_id
             )
+            pickle_path = os.path.join(
+                "/inference", request.project_id, f"embedder-{embedding_id}.pkl"
+            )
+            if not os.path.exists(pickle_path):
+                os.makedirs(os.path.dirname(pickle_path), exist_ok=True)
+                with open(pickle_path, "wb") as f:
+                    pickle.dump(embedder, f)
+
         upload_embedding_as_file(request.project_id, embedding_id)
         embedding.update_embedding_state_finished(
             request.project_id,
diff --git a/start b/start
index 3aedfd3..fdece59 100755
--- a/start
+++ b/start
@@ -27,6 +27,20 @@ then
     fi
 fi
 
+INFERENCE_DIR=${PWD%/*}/dev-setup/inference/
+if [ ! -d "$_DIR" ] 
+then
+    INFERENCE_DIR=${PWD%/*/*}/dev-setup/inference/
+    if [ ! -d "$INFERENCE_DIR" ] 
+    then
+        # to include volume for local development, use the dev-setup model data folder:  
+        # alternative use manual logic with
+        # -v /path/to/dev-setup/model-data:/models \
+        echo "Can't find model data directory: $INFERENCE_DIR -> stopping" 
+        exit 1
+    fi
+fi
+
 echo -ne 'starting...'
 docker run -d --rm \
 --name refinery-embedder \
@@ -42,6 +56,7 @@ docker run -d --rm \
 --mount type=bind,source="$(pwd)"/,target=/app \
 -v /var/run/docker.sock:/var/run/docker.sock \
 -v "$MODEL_DIR":/models \
+-v "$INFERENCE_DIR":/inference \
 --network dev-setup_default \
 refinery-embedder-dev > /dev/null 2>&1
 echo -ne '\t\t\t [done]\n'

From 65de4ff1ae3069317be33afde7f3b397bd896a42 Mon Sep 17 00:00:00 2001
From: felix0496 <felix.kirsch@kern.ai>
Date: Thu, 12 Jan 2023 14:00:35 +0100
Subject: [PATCH 2/4] write out embedder pickle for embedding type if is
 managed

---
 controller.py | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/controller.py b/controller.py
index b61834e..0abc036 100644
--- a/controller.py
+++ b/controller.py
@@ -450,6 +450,8 @@ def run_encoding(
             request_util.post_embedding_to_neural_search(
                 request.project_id, embedding_id
             )
+
+        if get_config_value("is_managed"):
             pickle_path = os.path.join(
                 "/inference", request.project_id, f"embedder-{embedding_id}.pkl"
             )

From 67ae661cfdaa0ded20f702df8872e97f8a6efc9a Mon Sep 17 00:00:00 2001
From: felix0496 <felix.kirsch@kern.ai>
Date: Mon, 16 Jan 2023 16:12:36 +0100
Subject: [PATCH 3/4] removes inference pickle on embedder removal

---
 controller.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/controller.py b/controller.py
index 0abc036..78c5369 100644
--- a/controller.py
+++ b/controller.py
@@ -496,6 +496,12 @@ def delete_embedding(project_id: str, embedding_id: str) -> int:
     org_id = organization.get_id_by_project_id(project_id)
     s3.delete_object(org_id, project_id + "/" + object_name)
     request_util.delete_embedding_from_neural_search(embedding_id)
+    if get_config_value("is_managed"):
+        pickle_path = os.path.join(
+            "/inference", project_id, f"embedder-{embedding_id}.pkl"
+        )
+        if os.path.exists(pickle_path):
+            os.remove(pickle_path)
     return 200
 
 

From adff62227fa59810fa8294d91c3a167d7e8b4e0a Mon Sep 17 00:00:00 2001
From: felix0496 <felix.kirsch@kern.ai>
Date: Mon, 30 Jan 2023 16:37:00 +0100
Subject: [PATCH 4/4] pr comments

---
 controller.py | 9 +++------
 start         | 4 ++--
 2 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/controller.py b/controller.py
index 78c5369..cc81153 100644
--- a/controller.py
+++ b/controller.py
@@ -496,12 +496,9 @@ def delete_embedding(project_id: str, embedding_id: str) -> int:
     org_id = organization.get_id_by_project_id(project_id)
     s3.delete_object(org_id, project_id + "/" + object_name)
     request_util.delete_embedding_from_neural_search(embedding_id)
-    if get_config_value("is_managed"):
-        pickle_path = os.path.join(
-            "/inference", project_id, f"embedder-{embedding_id}.pkl"
-        )
-        if os.path.exists(pickle_path):
-            os.remove(pickle_path)
+    pickle_path = os.path.join("/inference", project_id, f"embedder-{embedding_id}.pkl")
+    if os.path.exists(pickle_path):
+        os.remove(pickle_path)
     return 200
 
 
diff --git a/start b/start
index fdece59..e0dcd48 100755
--- a/start
+++ b/start
@@ -33,9 +33,9 @@ then
     INFERENCE_DIR=${PWD%/*/*}/dev-setup/inference/
     if [ ! -d "$INFERENCE_DIR" ] 
     then
-        # to include volume for local development, use the dev-setup model data folder:  
+        # to include volume for local development, use the dev-setup inference folder:  
         # alternative use manual logic with
-        # -v /path/to/dev-setup/model-data:/models \
+        # -v /path/to/dev-setup/inference:/models \
         echo "Can't find model data directory: $INFERENCE_DIR -> stopping" 
         exit 1
     fi