Merge pull request #79 from fractalego/removing-local-llm

Removing local llm

requirements.txt

@@ -1,11 +1,6 @@
 flask[async]==2.0.1
 flask-cors==3.0.10
 flask_dropzone==1.6.0
-accelerate==0.20.3
-bitsandbytes==0.41.3
-torch==2.0.0
-optimum==1.15.0
-transformers==4.35.2
 nltk==3.6.2
 gensim==4.3.1
 sklearn==0.0
@@ -21,6 +16,5 @@ sphinx==6.1.3
 sphinx-rtd-theme==1.2.0
 bluepy==1.3.0
 einops==0.6.1
-sentence_transformers==2.2.2
-fairseq==0.12.2
-g2p-en==2.1.0
+g2p-en==2.1.0
+pyyaml==6.0.1

setup.py

@@ -19,9 +19,7 @@
         "wafl.connectors",
         "wafl.connectors.bridges",
         "wafl.connectors.factories",
-        "wafl.connectors.local",
         "wafl.connectors.remote",
-        "wafl.entailment",
         "wafl.events",
         "wafl.extractors",
         "wafl.filter",
@@ -50,11 +48,6 @@
         "flask-cors==3.0.10",
         "flask_dropzone==1.6.0",
         "werkzeug==2.1.2",
-        "accelerate==0.20.3",
-        "bitsandbytes==0.41.3",
-        "torch==2.0.0",
-        "optimum==1.15.0",
-        "transformers==4.35.2",
         "nltk==3.6.2",
         "gensim==4.3.1",
         "sklearn==0.0",
@@ -67,9 +60,8 @@
         "word2number==1.1",
         "aiohttp==3.8.4",
         "einops==0.6.1",
-        "sentence_transformers==2.2.2",
-        "fairseq==0.12.2",
         "g2p-en==2.1.0",
+        "pyyaml==6.0.1",
     ],
     classifiers=[
         "License :: OSI Approved :: MIT License",

todo.txt

@@ -1,5 +1,8 @@
 ### TODO
 
+* no local llms, only remote
+* make it easy to run the llm on the server (something more than docker perhaps)?
+
 /* re-train the whisper model using the distilled version
 /* make rules reloadable
 /* nicer UI?

wafl/answerer/dialogue_answerer.py

@@ -13,7 +13,6 @@
 from wafl.connectors.bridges.llm_chitchat_answer_bridge import LLMChitChatAnswerBridge
 from wafl.exceptions import CloseConversation
 from wafl.extractors.dataclasses import Query, Answer
-from wafl.inference.utils import cluster_facts
 from wafl.simple_text_processing.questions import is_question
 
 

wafl/command_line.py

@@ -61,7 +61,8 @@ def process_cli():
 
         else:
             print("Unknown argument.\n")
-            print_help()
+    else:
+        print_help()
 
 
 def main():

wafl/connectors/base_llm_connector.py

@@ -1,13 +1,11 @@
 import logging
 import re
-import torch
 
 
 from wafl.connectors.utils import select_best_answer
 
 _system_logger = logging.getLogger(__file__)
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
 model = None
 tokenizer = None
 

wafl/connectors/factories/llm_connector_factory.py

@@ -1,11 +1,7 @@
-from wafl.connectors.local.local_llm_connector import LocalLLMConnector
 from wafl.connectors.remote.remote_llm_connector import RemoteLLMConnector
 
 
 class LLMConnectorFactory:
     @staticmethod
     def get_connector(config):
-        if config.get_value("llm_model")["model_is_local"]:
-            return LocalLLMConnector(config.get_value("llm_model"))
-
         return RemoteLLMConnector(config.get_value("llm_model"))

wafl/connectors/factories/sentence_embedder_connector_factory.py

@@ -1,6 +1,3 @@
-from wafl.connectors.local.local_sentence_embedder_connector import (
-    LocalSentenceEmbedderConnector,
-)
 from wafl.connectors.remote.remote_sentence_embedder_connector import (
     RemoteSentenceEmbedderConnector,
 )
@@ -9,9 +6,4 @@
 class SentenceEmbedderConnectorFactory:
     @staticmethod
     def get_connector(model_name, config):
-        if config.get_value(model_name)["model_is_local"]:
-            return LocalSentenceEmbedderConnector(
-                config.get_value(model_name)["local_model"]
-            )
-
         return RemoteSentenceEmbedderConnector(config.get_value(model_name))

wafl/connectors/factories/speaker_connector_factory.py

@@ -1,11 +1,7 @@
-from wafl.connectors.local.local_speaker_connector import LocalSpeakerConnector
 from wafl.connectors.remote.remote_speaker_connector import RemoteSpeakerConnector
 
 
 class SpeakerConnectorFactory:
     @staticmethod
     def get_connector(config):
-        if config.get_value("speaker_model")["model_is_local"]:
-            return LocalSpeakerConnector(config.get_value("speaker_model"))
-
-        return RemoteSpeakerConnector(config.get_value("speaker_model")["remote_model"])
+        return RemoteSpeakerConnector(config.get_value("speaker_model"))

wafl/connectors/factories/whisper_connector_factory.py

@@ -1,13 +1,9 @@
-from wafl.connectors.local.local_whisper_connector import LocalWhisperConnector
 from wafl.connectors.remote.remote_whisper_connector import RemoteWhisperConnector
 
 
 class WhisperConnectorFactory:
     @staticmethod
     def get_connector(config):
-        if config.get_value("listener_model")["model_is_local"]:
-            return LocalWhisperConnector(config.get_value("listener_model"))
-
         return RemoteWhisperConnector(
-            config.get_value("listener_model")["remote_model"]
+            config.get_value("listener_model")
         )