Merge pull request #274 from ChatWithPDF/restructure

changes to 4
Samagra-Development · Nov 9, 2023 · 1b58a99 · 1b58a99
2 parents 6667d8a + dc38d3a
commit 1b58a99
Show file tree

Hide file tree

Showing 6 changed files with 56 additions and 80 deletions.
diff --git a/config.json b/config.json
@@ -1,19 +1,27 @@
 {
   "models": [
+     {
+      "serviceName": "convo_starter",
+      "modelBasePath": "src/text_classification/convo_starter_orgbot/local/.",  
+      "apiBasePath": "text_classification/convo_starter_orgbot/local/",
+      "containerPort": 8000,
+      "environment": { },
+      "nginx": [],
+      "build": false
+    },
     {
       "serviceName": "search_tf",
       "modelBasePath": "src/search/tf_search/local/.",  
       "apiBasePath": "search/tf_search/local/",
       "containerPort": 8000,
       "environment": { },
       "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
-      "build": true
+      "build": false
     },
-
     {
-      "serviceName": "embedding_instructor_gpu",
-      "modelBasePath": "src/embeddings/instructor_gpu/local/.",
-      "apiBasePath": "embeddings/instructor_gpu/local/",
+      "serviceName": "embedding_instructor",
+      "modelBasePath": "src/embeddings/instructor/local/.",
+      "apiBasePath": "/embeddings/instructor/local",
       "containerPort": 8000,
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}",
@@ -23,6 +31,17 @@
       "constraints": ["node.labels.node_vm_type==gpu"],
       "build": false
     },
+    {
+      "serviceName": "embedding_openai",
+      "modelBasePath": "src/embeddings/openai/remote/.",
+      "apiBasePath": "/embeddings/openai/remote",
+      "containerPort": 8000,
+      "environment": {
+        "OPENAI_API_KEY": "${OPENAI_API_KEY}"
+      },
+      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
+      "build": false
+    },
     {
       "serviceName": "asr_whisper_en",
       "modelBasePath": "src/asr/whisper_en/local/.",
@@ -72,16 +91,6 @@
       "nginx": [],
       "build": false
     },
-    {
-      "serviceName": "flow_classification",
-      "modelBasePath": "src/text_classification/flow_classification/local/.",
-      "apiBasePath": "text_classification/flow_classification/local/",
-      "containerPort": 8000,
-      "environment": {},
-      "nginx": [],
-      "constraints": ["node.role==worker"],
-      "build": false
-    },
     {
       "serviceName": "text_translation_azure_dict",
       "modelBasePath": "src/text_translation/azure_dict/remote/.",
@@ -178,27 +187,26 @@
       "build": false
     },
     {
-      "serviceName": "embedding_instructor",
-      "modelBasePath": "src/embeddings/instructor/local/.",
-      "apiBasePath": "/embeddings/instructor/local",
+      "serviceName": "llm_openai_gpt3",
+      "modelBasePath": "src/llm/openai/chatgpt3/.",
+      "apiBasePath": "/llm/openai/chatgpt3",
       "containerPort": 8000,
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
-      "nginx": ["client_max_body_size 100M;", "proxy_read_timeout 600;", "proxy_connect_timeout 600;", "proxy_send_timeout 600;"],
+      "nginx": [],
       "constraints": ["node.labels.node_vm_type==gpu"],
       "build": false
     },
     {
-      "serviceName": "llm_openai_gpt3",
-      "modelBasePath": "src/llm/openai/chatgpt3/.",
-      "apiBasePath": "/llm/openai/chatgpt3",
+      "serviceName": "llm_openai_gpt4",
+      "modelBasePath": "src/llm/openai/chatgpt4/.",
+      "apiBasePath": "/llm/openai/chatgpt4",
       "containerPort": 8000,
       "environment": {
         "OPENAI_API_KEY": "${OPENAI_API_KEY}"
       },
       "nginx": [],
-      "constraints": ["node.labels.node_vm_type==gpu"],
       "build": true
     }
   ]

diff --git a/src/embeddings/openai/remote/model.py b/src/embeddings/openai/remote/model.py
@@ -1,52 +1,28 @@
-import os
-import openai
-from openai.embeddings_utils import get_embedding
-from cache import AsyncTTL
 from request import ModelRequest
-import numpy as np
-import pandas as pd
-import tiktoken
-import ast
-from sklearn.metrics.pairwise import cosine_similarity
-
-openai.api_key = os.getenv("OPENAI_API_KEY")
-
+from openai import OpenAI
+import os
 
 class Model:
-    embedding_df = None
     embedding_model = "text-embedding-ada-002"
-    embedding_encoding = "cl100k_base"  # this the encoding for text-embedding-ada-002
-    max_tokens = 8000  # the maximum for text-embedding-ada-002 is 8191
 
     def __new__(cls, context):
         cls.context = context
         if not hasattr(cls, 'instance'):
-            cls.embedding_df = pd.read_csv('src/embeddings/openai/remote/akai.csv')
-            cls.embedding_df['embedding'] = cls.embedding_df['embedding'].apply(ast.literal_eval)
+            cls.client = OpenAI(
+                api_key=os.getenv("OPENAI_API_KEY"),
+            )
             cls.instance = super(Model, cls).__new__(cls)
         return cls.instance
 
-    @AsyncTTL(time_to_live=600000, maxsize=1024)
     async def inference(self, request: ModelRequest):
-        print("request.prompt", request.prompt)
-        new_prompt_embedding = get_embedding(request.prompt, engine=self.embedding_model)
-        similarity_scores = cosine_similarity(
-            [new_prompt_embedding], np.stack(self.embedding_df['embedding'], axis=0))[0]
-        most_similar_indices = np.argsort(similarity_scores)[::-1]
-        most_similar_prompts = self.embedding_df.loc[most_similar_indices, ['combined_prompt', 'combined_content']]
-        most_similar_prompts['similarity_score'] = np.sort(similarity_scores)[::-1]
-        similar_content = most_similar_prompts.iloc[0:20]
-        sim_cutoff_range = np.max(similar_content['similarity_score']) - request.similarity_score_range
-        similar_content_df = similar_content.loc[similar_content['similarity_score'] >= sim_cutoff_range, :]
-        similar_content_df1 = similar_content_df.drop(columns='similarity_score')
-        similar_content_dict = similar_content_df1.to_dict('records')
-        # modified_content_dict = remove_content_tags_from_dic(similar_content_dict)
-        print("similar_content_dict", similar_content_dict)
-        return (similar_content_dict)
+    # Modify this function according to model requirements such that inputs and output remains the same
+        query = request.query
 
-    async def create_embeddings(self, embedding_df):
-        encoding = tiktoken.get_encoding(self.embedding_encoding)
-        embedding_df["n_tokens"] = embedding_df.combined_prompt.apply(lambda x: len(encoding.encode(x)))
-        embedding_df["embedding"] = embedding_df.combined_prompt.apply(
-            lambda x: get_embedding(x, engine=self.embedding_model))
-        return embedding_df
+        if(query != None):
+            embedding = self.client.embeddings.create(
+                input=query,
+                model=self.embedding_model,
+            ).data[0].embedding
+            return [embedding]
+
+        return "Invalid input"
diff --git a/src/embeddings/openai/remote/request.py b/src/embeddings/openai/remote/request.py
@@ -1,10 +1,11 @@
 import json
 
-
 class ModelRequest():
-    def __init__(self, prompt, similarity_score_range=0):
-        self.prompt = prompt
-        self.similarity_score_range = similarity_score_range
+    def __init__(self, query=None, df = None, query_type =  None):
+         # Url to download csv file
+        self.query = query # String
+        self.query_type =  query_type
+        self.df = df
 
     def to_json(self):
         return json.dumps(self, default=lambda o: o.__dict__,

diff --git a/src/embeddings/openai/remote/requirements.txt b/src/embeddings/openai/remote/requirements.txt
@@ -1,9 +1,5 @@
 aiohttp==3.8.4
-quart==0.18.3
+quart
 async-cache==1.1.1
 requests
 openai
-numpy
-pandas
-tiktoken
-sklearn
diff --git a/src/llm/openai/chatgpt4/model.py b/src/llm/openai/chatgpt4/model.py
@@ -15,7 +15,6 @@ def __new__(cls, context):
             cls.instance = super(Model, cls).__new__(cls)
         return cls.instance
 
-    @AsyncTTL(time_to_live=600000, maxsize=1024)
     @retry(wait=wait_random_exponential(min=1, max=20), stop=stop_after_attempt(6))
     async def inference(self, request: ModelRequest):
         response = await openai_async.chat_complete(
@@ -24,11 +23,7 @@ async def inference(self, request: ModelRequest):
             payload={
                 "model": "gpt-4",
                 "temperature": 0,
-                "messages": [{"role":"user","content" : request.prompt}],
+                "messages": request.prompt,
             },
         )
-        try:
-            ans = response.json()["choices"][0]["message"]["content"]
-            return {"ans":ans}
-        except:
-            return response.json()
+        return response.json()
diff --git a/src/llm/openai/chatgpt4/requirements.txt b/src/llm/openai/chatgpt4/requirements.txt
@@ -1,7 +1,7 @@
 aiohttp==3.8.4
-quart==0.18.3
+quart
 async-cache==1.1.1
 requests
 openai
 openai_async
-tenacity
+tenacity