gooberinc/goober
3
0
Fork 1
Code Issues 1 Pull requests Projects Releases Packages Wiki Activity Actions

lobotmized tf.py to work with metal

Browse source
This commit is contained in:
WhatDidYouExpect 2025-01-06 00:23:59 +01:00 • committed by GitHub
parent 5dfe6449cb
commit c84faaed50
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 48 additions and 51 deletions
Download patch file Download diff file Expand all files Collapse all files

99
customcommands/tf.py
View file

@ -7,27 +7,23 @@ import json
from time import strftime, localtime from time import strftime, localtime
import pickle import pickle
import re import re
from discord import app_commands
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, LSTM, Dense
from tensorflow.keras.models import load_model
from tensorflow.keras.backend import clear_session
ready: bool = True ready: bool = True
MODEL_MATCH_STRING = "[0-9]{2}_[0-9]{2}_[0-9]{4}-[0-9]{2}_[0-9]{2}" MODEL_MATCH_STRING = "[0-9]{2}_[0-9]{2}_[0-9]{4}-[0-9]{2}_[0-9]{2}"
try: try:
import tensorflow as tf tf.config.optimizer.set_jit(False)
from tensorflow import keras
from keras.preprocessing.text import Tokenizer
from keras_preprocessing.sequence import pad_sequences
from keras.models import Sequential
from keras.layers import Embedding, LSTM, Dense
from keras.models import load_model
from keras.backend import clear_session
tf.config.optimizer.set_jit(True)
except ImportError: except ImportError:
print("ERROR: Failed to import Tensorflow. Here is a list of required dependencies:",( print("ERROR: Failed to import TensorFlow.")
"tensorflow==2.10.0"
"(for Nvidia users: tensorflow-gpu==2.10.0)"
"(for macOS: tensorflow-metal==0.6.0, tensorflow-macos==2.10.0)"
"numpy~=1.23"
))
ready = False ready = False
class Ai: class Ai:
@ -36,7 +32,7 @@ class Ai:
if model_path: if model_path:
self.__load_model(model_path) self.__load_model(model_path)
self.is_loaded = model_path is not None self.is_loaded = model_path is not None
self.batch_size = 64 self.batch_size = 32
def get_model_name_from_path(self,path:str): def get_model_name_from_path(self,path:str):
print(path) print(path)
@ -88,7 +84,7 @@ class Ai:
self.tokenizer = Tokenizer() self.tokenizer = Tokenizer()
with open("memory.json","r") as f: with open("memory.json","r") as f:
self.tokenizer.fit_on_sequences(json.load(f)) self.tokenizer.fit_on_texts(json.load(f))
self.is_loaded = True self.is_loaded = True
def reload_model(self): def reload_model(self):
@ -144,7 +140,7 @@ class Learning(Ai):
x_pad = pad_sequences(X, maxlen=maxlen, padding="pre") x_pad = pad_sequences(X, maxlen=maxlen, padding="pre")
y = np.array(y) y = np.array(y)
history = self.model.fit(x_pad,y, epochs=iters, validation_data=(x_pad,y), batch_size=64) # Idelaly, validation data would be seperate from the actual data history = self.model.fit(x_pad,y, epochs=iters, validation_data=(x_pad,y), batch_size=64) # Ideally, validation data would be separate from the actual data
self.save_model(self.model,tokenizer,history,self.get_model_name_from_path(settings.get("model_path"))) self.save_model(self.model,tokenizer,history,self.get_model_name_from_path(settings.get("model_path")))
class Generation(Ai): class Generation(Ai):
@ -247,53 +243,54 @@ class Tf(commands.Cog):
return inner return inner
def __init__(self,bot): def __init__(self, bot):
global learning, generation global learning, generation
global ready global ready
os.makedirs(os.path.join(".","models"),exist_ok=True) os.makedirs(os.path.join(".","models"), exist_ok=True)
Settings().load() Settings().load()
self.bot = bot self.bot = bot
learning = Learning() learning = Learning()
generation = Generation() generation = Generation()
@commands.command() @app_commands.command(name="start", description="Starts the bot")
async def start(self,ctx): async def start(self, interaction: discord.Interaction):
await ctx.defer() await interaction.response.send_message("hi")
await ctx.send("hi")
@commands.command() @app_commands.command(name="generate", description="Generates a sentence")
async def generate(self,ctx,seed:str,word_amount:int=5): async def generate(self, interaction: discord.Interaction, seed: str, word_amount: int = 5):
await ctx.defer() await interaction.response.defer()
await ctx.send(generation.generate_sentence(word_amount,seed)) sentence = generation.generate_sentence(word_amount, seed)
await interaction.followup.send(sentence)
@commands.command() @app_commands.command(name="create", description="Trains the model with memory")
async def create(self,ctx): async def create(self, interaction: discord.Interaction):
await ctx.defer() await interaction.response.defer()
with open("memory.json","r") as f: with open("memory.json", "r") as f:
memory:List[str] = json.load(f) memory: List[str] = json.load(f)
learning.create_model(memory) # TODO: CHANGE learning.create_model(memory) # TODO: CHANGE
await ctx.send("Trained succesfully!") await interaction.followup.send("Trained successfully!")
@commands.command() @app_commands.command(name="train", description="Trains the model further with memory")
async def train(self,ctx): async def train(self, interaction: discord.Interaction):
await ctx.defer() await interaction.response.defer()
with open("memory.json","r") as f: with open("memory.json", "r") as f:
memory:List[str] = json.load(f) memory: List[str] = json.load(f)
learning.add_training(memory,2) learning.add_training(memory, 2)
await ctx.send("Finished!") await interaction.followup.send("Finished training!")
@commands.command() @app_commands.command(name="change", description="Change the model")
async def change(self,ctx,model:str=None): async def change(self, interaction: discord.Interaction, model: str = None):
embed = discord.Embed(title="Change model",description="Which model would you like to use?") embed = discord.Embed(title="Change model", description="Which model would you like to use?")
if model is None: if model is None:
models:List[str] = os.listdir(os.path.join(".","models")) models: List[str] = os.listdir(os.path.join(".", "models"))
models = [folder for folder in models if re.match(MODEL_MATCH_STRING,folder)] models = [folder for folder in models if re.match(MODEL_MATCH_STRING, folder)]
if len(models) == 0: if len(models) == 0:
models = ["No models available."] models = ["No models available."]
await ctx.send(embed=embed,view=DropdownView(90,models)) await interaction.response.send_message(embed=embed, view=DropdownView(90, models))
learning.reload_model() learning.reload_model()
generation.reload_model() generation.reload_model()
async def setup(bot): async def setup(bot):
await bot.add_cog(Tf(bot)) await bot.add_cog(Tf(bot))