YOURNAME
commited on
Commit
·
e09c84c
1
Parent(s):
4870f5c
- pyproject.toml +2 -8
- src/pipeline.py +59 -53
pyproject.toml
CHANGED
@@ -23,20 +23,14 @@ dependencies = [
|
|
23 |
]
|
24 |
|
25 |
[[tool.edge-maxxing.models]]
|
26 |
-
repository = "
|
27 |
-
revision = "
|
28 |
|
29 |
[[tool.edge-maxxing.models]]
|
30 |
repository = "city96/t5-v1_1-xxl-encoder-bf16"
|
31 |
revision = "1b9c856aadb864af93c1dcdc226c2774fa67bc86"
|
32 |
|
33 |
-
[[tool.edge-maxxing.models]]
|
34 |
-
repository = "MyApricity/Vae_Only"
|
35 |
-
revision = "a47d57702caf8ff0c0e21d30b93f9d3297b81920"
|
36 |
|
37 |
-
[[tool.edge-maxxing.models]]
|
38 |
-
repository = "MyApricity/Flux_Transformer_float8"
|
39 |
-
revision = "66c5f182385555a00ec90272ab711bb6d3c197db"
|
40 |
|
41 |
[project.scripts]
|
42 |
start_inference = "main:main"
|
|
|
23 |
]
|
24 |
|
25 |
[[tool.edge-maxxing.models]]
|
26 |
+
repository = "MyApricity/FLUX_OPT_SCHNELL_1.2"
|
27 |
+
revision = "488528b6f815bff1bbc747cf1e0947c77c544665"
|
28 |
|
29 |
[[tool.edge-maxxing.models]]
|
30 |
repository = "city96/t5-v1_1-xxl-encoder-bf16"
|
31 |
revision = "1b9c856aadb864af93c1dcdc226c2774fa67bc86"
|
32 |
|
|
|
|
|
|
|
33 |
|
|
|
|
|
|
|
34 |
|
35 |
[project.scripts]
|
36 |
start_inference = "main:main"
|
src/pipeline.py
CHANGED
@@ -8,7 +8,8 @@ import transformers
|
|
8 |
from huggingface_hub.constants import HF_HUB_CACHE
|
9 |
from transformers import T5EncoderModel, T5TokenizerFast, CLIPTokenizer, CLIPTextModel
|
10 |
|
11 |
-
|
|
|
12 |
from torch import Generator
|
13 |
from diffusers import FluxTransformer2DModel, DiffusionPipeline
|
14 |
|
@@ -19,60 +20,49 @@ from optimum.quanto import requantize
|
|
19 |
import json
|
20 |
|
21 |
|
|
|
22 |
|
23 |
|
24 |
torch._dynamo.config.suppress_errors = True
|
25 |
os.environ['PYTORCH_CUDA_ALLOC_CONF']="expandable_segments:True"
|
26 |
os.environ["TOKENIZERS_PARALLELISM"] = "True"
|
27 |
|
28 |
-
|
29 |
-
|
30 |
Pipeline = None
|
31 |
-
|
32 |
|
33 |
import torch
|
34 |
import math
|
35 |
from typing import Dict, Any
|
36 |
|
37 |
def remove_cache():
|
38 |
-
gc.collect()
|
39 |
torch.cuda.empty_cache()
|
40 |
torch.cuda.reset_max_memory_allocated()
|
|
|
41 |
torch.cuda.reset_peak_memory_stats()
|
42 |
|
43 |
|
44 |
-
class InitializingModel:
|
45 |
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
return text_encoder.to(memory_format=torch.channels_last)
|
55 |
|
56 |
-
@staticmethod
|
57 |
-
def load_transformer(trans_path: str) -> FluxTransformer2DModel:
|
58 |
-
print("Loading transformer model...")
|
59 |
-
transformer = FluxTransformer2DModel.from_pretrained(
|
60 |
-
trans_path,
|
61 |
-
torch_dtype=torch.bfloat16,
|
62 |
-
use_safetensors=False,
|
63 |
-
)
|
64 |
-
return transformer.to(memory_format=torch.channels_last)
|
65 |
|
66 |
-
|
67 |
-
class CompileTransformerDiffusion:
|
68 |
def __init__(self, pipeline, optimize=False):
|
69 |
self.pipeline = pipeline
|
70 |
self.optimize = optimize
|
71 |
if self.optimize:
|
72 |
-
self.
|
73 |
|
74 |
-
def
|
75 |
-
|
76 |
self.pipeline.unet = torch.compile(self.pipeline.unet)
|
77 |
|
78 |
def __call__(self, *args, **kwargs):
|
@@ -80,39 +70,55 @@ class CompileTransformerDiffusion:
|
|
80 |
|
81 |
def load_pipeline() -> Pipeline:
|
82 |
|
|
|
83 |
|
84 |
-
|
85 |
-
base_transformer = InitializingModel.load_transformer(base_transformer_path)
|
86 |
|
87 |
-
|
88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
-
pipeline = DiffusionPipeline.from_pretrained(CHECKPOINT,
|
91 |
-
revision=REVISION,
|
92 |
-
transformer=base_transformer,
|
93 |
-
text_encoder_2=text_encoder_2,
|
94 |
-
torch_dtype=torch.bfloat16)
|
95 |
pipeline.to("cuda")
|
|
|
96 |
try:
|
97 |
-
pipeline
|
98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
except:
|
100 |
-
print("
|
101 |
|
102 |
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
|
|
|
|
116 |
|
117 |
return pipeline
|
118 |
|
|
|
8 |
from huggingface_hub.constants import HF_HUB_CACHE
|
9 |
from transformers import T5EncoderModel, T5TokenizerFast, CLIPTokenizer, CLIPTextModel
|
10 |
|
11 |
+
# ApricityApricityApricityApricityApricityApricityApricityApricityApricityApricityApricity
|
12 |
+
|
13 |
from torch import Generator
|
14 |
from diffusers import FluxTransformer2DModel, DiffusionPipeline
|
15 |
|
|
|
20 |
import json
|
21 |
|
22 |
|
23 |
+
# ApricityApricityApricityApricityApricityApricityApricityApricityApricityApricityApricity
|
24 |
|
25 |
|
26 |
torch._dynamo.config.suppress_errors = True
|
27 |
os.environ['PYTORCH_CUDA_ALLOC_CONF']="expandable_segments:True"
|
28 |
os.environ["TOKENIZERS_PARALLELISM"] = "True"
|
29 |
|
30 |
+
ckpt_root = "MyApricity/FLUX_OPT_SCHNELL_1.2"
|
31 |
+
revision_root = "488528b6f815bff1bbc747cf1e0947c77c544665"
|
32 |
Pipeline = None
|
33 |
+
use_com = False
|
34 |
|
35 |
import torch
|
36 |
import math
|
37 |
from typing import Dict, Any
|
38 |
|
39 |
def remove_cache():
|
|
|
40 |
torch.cuda.empty_cache()
|
41 |
torch.cuda.reset_max_memory_allocated()
|
42 |
+
gc.collect()
|
43 |
torch.cuda.reset_peak_memory_stats()
|
44 |
|
45 |
|
|
|
46 |
|
47 |
+
def text_t5_loader() -> T5EncoderModel:
|
48 |
+
print("Loading text encoder...")
|
49 |
+
text_encoder = T5EncoderModel.from_pretrained(
|
50 |
+
"city96/t5-v1_1-xxl-encoder-bf16",
|
51 |
+
revision="1b9c856aadb864af93c1dcdc226c2774fa67bc86",
|
52 |
+
torch_dtype=torch.bfloat16,
|
53 |
+
)
|
54 |
+
return text_encoder.to(memory_format=torch.channels_last)
|
|
|
55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
|
57 |
+
class StableDiffusionTransformerCompile:
|
|
|
58 |
def __init__(self, pipeline, optimize=False):
|
59 |
self.pipeline = pipeline
|
60 |
self.optimize = optimize
|
61 |
if self.optimize:
|
62 |
+
self.model_compiling()
|
63 |
|
64 |
+
def model_compiling(self):
|
65 |
+
# Staff doing here
|
66 |
self.pipeline.unet = torch.compile(self.pipeline.unet)
|
67 |
|
68 |
def __call__(self, *args, **kwargs):
|
|
|
70 |
|
71 |
def load_pipeline() -> Pipeline:
|
72 |
|
73 |
+
text_t5_encoder = text_t5_loader()
|
74 |
|
75 |
+
transformer_path__ = os.path.join(HF_HUB_CACHE, "models--MyApricity--FLUX_OPT_SCHNELL_1.2/snapshots/488528b6f815bff1bbc747cf1e0947c77c544665")
|
|
|
76 |
|
77 |
+
transformer__ = FluxTransformer2DModel.from_pretrained(transformer_path__, torch_dtype=torch.bfloat16, use_safetensors=False)
|
78 |
|
79 |
+
try:
|
80 |
+
pipeline = DiffusionPipeline.from_pretrained(ckpt_root,
|
81 |
+
revision=revision_root,
|
82 |
+
transformer=transformer__,
|
83 |
+
torch_dtype=torch.bfloat16)
|
84 |
+
|
85 |
+
except:
|
86 |
+
pipeline = DiffusionPipeline.from_pretrained(ckpt_root,
|
87 |
+
revision=revision_root,
|
88 |
+
torch_dtype=torch.bfloat16)
|
89 |
|
|
|
|
|
|
|
|
|
|
|
90 |
pipeline.to("cuda")
|
91 |
+
|
92 |
try:
|
93 |
+
compiled_pipeline = StableDiffusionTransformerCompile(pipeline, optimize=False)
|
94 |
+
|
95 |
+
if use_com:
|
96 |
+
pipeline = compiled_pipeline
|
97 |
+
else:
|
98 |
+
print("Currently not compling affectively")
|
99 |
+
|
100 |
+
pipeline.disable_vae_compress()
|
101 |
+
pipeline.text_encoder_2 = text_t5_encoder
|
102 |
+
|
103 |
except:
|
104 |
+
print("pipeline")
|
105 |
|
106 |
|
107 |
+
prompt_1 = "albaspidin, pillmonger, palaeocrystalline"
|
108 |
+
pipeline(prompt=prompt_1,
|
109 |
+
width=1024,
|
110 |
+
height=1024,
|
111 |
+
guidance_scale=0.0,
|
112 |
+
num_inference_steps=4,
|
113 |
+
max_sequence_length=256)
|
114 |
+
|
115 |
+
prompt_2 = "obe, kilometrage, circuition"
|
116 |
+
pipeline(prompt=prompt_2,
|
117 |
+
width=1024,
|
118 |
+
height=1024,
|
119 |
+
guidance_scale=0.0,
|
120 |
+
num_inference_steps=4,
|
121 |
+
max_sequence_length=256)
|
122 |
|
123 |
return pipeline
|
124 |
|