enhanced diffusers

This commit is contained in:
Saifeddine ALOUI 2024-05-27 23:59:55 +02:00
parent dad67273b5
commit 7468f4414c
6 changed files with 77 additions and 15 deletions

View File

@ -1,5 +1,5 @@
# =================== Lord Of Large Language Multimodal Systems Configuration file =========================== # =================== Lord Of Large Language Multimodal Systems Configuration file ===========================
version: 102 version: 104
binding_name: null binding_name: null
model_name: null model_name: null
model_variant: null model_variant: null
@ -151,6 +151,10 @@ openai_tts_voice: "alloy"
enable_sd_service: false enable_sd_service: false
sd_base_url: http://localhost:7860 sd_base_url: http://localhost:7860
# diffuser
diffusers_offloading_mode: sequential_cpu_offload # sequential_cpu_offload
diffusers_model: PixArt-alpha/PixArt-Sigma-XL-2-1024-MS
# Dall e service key # Dall e service key
dall_e_key: "" dall_e_key: ""
dall_e_generation_engine: "dall-e-3" dall_e_generation_engine: "dall-e-3"

View File

@ -1,5 +1,5 @@
# =================== Lord Of Large Language Multimodal Systems Configuration file =========================== # =================== Lord Of Large Language Multimodal Systems Configuration file ===========================
version: 102 version: 104
binding_name: null binding_name: null
model_name: null model_name: null
model_variant: null model_variant: null
@ -151,6 +151,10 @@ openai_tts_voice: "alloy"
enable_sd_service: false enable_sd_service: false
sd_base_url: http://localhost:7860 sd_base_url: http://localhost:7860
# diffuser
diffusers_offloading_mode: sequential_cpu_offload # sequential_cpu_offload
diffusers_model: PixArt-alpha/PixArt-Sigma-XL-2-1024-MS
# Dall e service key # Dall e service key
dall_e_key: "" dall_e_key: ""
dall_e_generation_engine: "dall-e-3" dall_e_generation_engine: "dall-e-3"

View File

@ -0,0 +1,38 @@
# Lollms function call definition file
# Here you need to import any necessary imports depending on the function requested by the user
import random
# Partial is useful if we need to preset some parameters
from functools import partial
# It is advised to import typing elements
# from typing import List
# Import PackageManager if there are potential libraries that need to be installed
from lollms.utilities import PackageManager
# ascii_colors offers advanced console coloring and bug tracing
from ascii_colors import trace_exception
# here is the core of the function to be built
def roll_a_dice() -> int:
try:
# handle exceptions
# Perform dice roll
result = random.randint(1, 6)
# Return the dice roll result
return result
except Exception as e:
return trace_exception(e)
#Here is the metadata function that should have the name in format function_name_function
def roll_a_dice_function():
return {
"function_name": "roll_a_dice", # The function name in string
"function": roll_a_dice, # The function to be called
"function_description": "Returns a random dice roll result between 1 and 6.", # Description of the function
"function_parameters": [] # No parameters needed for this function
}

View File

@ -1,5 +1,5 @@
# =================== Lord Of Large Language Multimodal Systems Configuration file =========================== # =================== Lord Of Large Language Multimodal Systems Configuration file ===========================
version: 102 version: 104
binding_name: null binding_name: null
model_name: null model_name: null
model_variant: null model_variant: null
@ -151,6 +151,10 @@ openai_tts_voice: "alloy"
enable_sd_service: false enable_sd_service: false
sd_base_url: http://localhost:7860 sd_base_url: http://localhost:7860
# diffuser
diffusers_offloading_mode: sequential_cpu_offload # sequential_cpu_offload
diffusers_model: PixArt-alpha/PixArt-Sigma-XL-2-1024-MS
# Dall e service key # Dall e service key
dall_e_key: "" dall_e_key: ""
dall_e_generation_engine: "dall-e-3" dall_e_generation_engine: "dall-e-3"

View File

@ -8,7 +8,7 @@ import sys
from lollms.app import LollmsApplication from lollms.app import LollmsApplication
from lollms.paths import LollmsPaths from lollms.paths import LollmsPaths
from lollms.config import TypedConfig, ConfigTemplate, BaseConfig from lollms.config import TypedConfig, ConfigTemplate, BaseConfig
from lollms.utilities import PackageManager, check_and_install_torch from lollms.utilities import PackageManager, check_and_install_torch, find_next_available_filename
import time import time
import io import io
import sys import sys
@ -73,14 +73,17 @@ def install_diffusers(lollms_app:LollmsApplication):
root_dir = lollms_app.lollms_paths.personal_path root_dir = lollms_app.lollms_paths.personal_path
shared_folder = root_dir/"shared" shared_folder = root_dir/"shared"
diffusers_folder = shared_folder / "diffusers" diffusers_folder = shared_folder / "diffusers"
diffusers_folder.mkdir(exist_ok=True, parents=True)
if not PackageManager.check_package_installed("diffusers"): if not PackageManager.check_package_installed("diffusers"):
PackageManager.install_or_update("diffusers") PackageManager.install_or_update("diffusers")
diffusers_folder.mkdir(exist_ok=True, parents=True) PackageManager.install_or_update("xformers")
def upgrade_diffusers(lollms_app:LollmsApplication): def upgrade_diffusers(lollms_app:LollmsApplication):
PackageManager.install_or_update("diffusers") PackageManager.install_or_update("diffusers")
PackageManager.install_or_update("xformers")
class LollmsDiffusers(LollmsTTI): class LollmsDiffusers(LollmsTTI):
@ -115,13 +118,17 @@ class LollmsDiffusers(LollmsTTI):
ASCIIColors.red(" ______ ") ASCIIColors.red(" ______ ")
ASCIIColors.red(" |______| ") ASCIIColors.red(" |______| ")
import torch import torch
from diffusers import PixArtSigmaPipeline from diffusers import PixArtSigmaPipeline
self.model = PixArtSigmaPipeline.from_pretrained( self.model = PixArtSigmaPipeline.from_pretrained(
"PixArt-alpha/PixArt-Sigma-XL-2-1024-MS", torch_dtype=torch.float16, cache_dir=self.models_dir app.config.diffusers_model, torch_dtype=torch.float16, cache_dir=self.models_dir,
use_safetensors=True,
) )
# Enable memory optimizations. # Enable memory optimizations.
self.model.enable_model_cpu_offload() if app.config.diffusers_offloading_mode=="sequential_cpu_offload":
self.model.enable_sequential_cpu_offload()
elif app.coinfig.diffusers_offloading_mode=="model_cpu_offload":
self.model.enable_model_cpu_offload()
@staticmethod @staticmethod
def verify(app:LollmsApplication): def verify(app:LollmsApplication):
@ -162,11 +169,13 @@ class LollmsDiffusers(LollmsTTI):
restore_faces=True, restore_faces=True,
output_path=None output_path=None
): ):
if output_path is None:
array = self.model(diffusers_positive_prompt).images[0] output_path = self.output_dir
image = Image.fromarray(array) from diffusers.utils.pil_utils import pt_to_pil
image = self.model(diffusers_positive_prompt, negative_prompt=diffusers_negative_prompt, guidance_scale=scale, num_inference_steps=steps,).images[0]
output_path = Path(output_path)
fn = find_next_available_filename(output_path,"diff_img_")
# Save the image # Save the image
image.save('output_image.png') image.save(fn)
return None, None return fn, {"prompt":diffusers_positive_prompt, "negative_prompt":diffusers_negative_prompt}

View File

@ -199,7 +199,10 @@ class LollmsXTTS(LollmsTTS):
"top_k": self.app.config.xtts_top_k, "top_k": self.app.config.xtts_top_k,
"enable_text_splitting": self.app.config.xtts_enable_text_splitting "enable_text_splitting": self.app.config.xtts_enable_text_splitting
} }
response = requests.post(f"{self.xtts_base_url}/set_tts_settings", settings) response = requests.post(f"{self.xtts_base_url}/set_tts_settings", settings,headers={
'accept': 'application/json',
'Content-Type': 'application/json'
})
if response.status_code == 200: if response.status_code == 200:
ASCIIColors.success("XTTS updated successfully") ASCIIColors.success("XTTS updated successfully")
except Exception as ex: except Exception as ex: