VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-18 01:30:02 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update accel_sdxl_gen_img.py

Browse Source
This commit is contained in:
DKnight54
2025-02-08 21:05:34 +08:00
committed by GitHub
parent 0105b920ab
commit f06d23cb1f
1 changed files with 14 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
accel_sdxl_gen_img.py
View File

@@ -1513,13 +1513,12 @@ def main(args):
logger.info(f"preferred device: {device}, {distributed_state.is_main_process}") logger.info(f"preferred device: {device}, {distributed_state.is_main_process}")
clean_memory_on_device(device) clean_memory_on_device(device)
model_dtype = sdxl_train_util.match_mixed_precision(args, dtype) model_dtype = sdxl_train_util.match_mixed_precision(args, dtype)
for pi in range(distributed_state.num_processes):
if pi == distributed_state.local_process_index: logger.info(f"loading model for process {distributed_state.local_process_index+1}/{distributed_state.num_processes}")
logger.info(f"loading model for process {distributed_state.local_process_index+1}/{distributed_state.num_processes}") (_, text_encoder1, text_encoder2, vae, unet, _, _) = sdxl_train_util._load_target_model(
(_, text_encoder1, text_encoder2, vae, unet, _, _) = sdxl_train_util._load_target_model( args.ckpt, args.vae, sdxl_model_util.MODEL_VERSION_SDXL_BASE_V1_0, dtype, device if args.lowram else "cpu", model_dtype
args.ckpt, args.vae, sdxl_model_util.MODEL_VERSION_SDXL_BASE_V1_0, dtype, device if args.lowram else "cpu", model_dtype )
) unet: InferSdxlUNet2DConditionModel = InferSdxlUNet2DConditionModel(unet)
unet: InferSdxlUNet2DConditionModel = InferSdxlUNet2DConditionModel(unet)
distributed_state.wait_for_everyone() distributed_state.wait_for_everyone()
# xformers、Hypernetwork対応 # xformers、Hypernetwork対応
@@ -2897,8 +2896,14 @@ def main(args):
res = [j for j, val in enumerate(prompt_data_list) if val.ext == unique_extinfo[i]] res = [j for j, val in enumerate(prompt_data_list) if val.ext == unique_extinfo[i]]
for index in res: for index in res:
templist.append(prompt_data_list[index]) templist.append(prompt_data_list[index])
if distributed_state.num_processes > 1:
resorted_list = []
for i in range(distributed_state.num_processes):
resorted_list.append(templist[i :: distributed_state.num_processes])
for list in resorted_list:
templist.extend(list)
split_into_batches = get_batches(items=templist, batch_size=args.batch_size).copy() split_into_batches = get_batches(items=templist, batch_size=args.batch_size).copy()
'''
sublist = [] sublist = []
if(len(split_into_batches) % distributed_state.num_processes != 0): if(len(split_into_batches) % distributed_state.num_processes != 0):
#Distributes last round of batches across available processes if last round of batches less than available processes and there is more than one prompt in the last batch #Distributes last round of batches across available processes if last round of batches less than available processes and there is more than one prompt in the last batch
@@ -2917,7 +2922,7 @@ def main(args):
sublist.reverse() sublist.reverse()
n, m = divmod(len(sublist), distributed_state.num_processes) n, m = divmod(len(sublist), distributed_state.num_processes)
split_into_batches.extend([sublist[i*n+min(i,m):(i+1)*n+min(i+1,m)] for i in range(distributed_state.num_processes)]) split_into_batches.extend([sublist[i*n+min(i,m):(i+1)*n+min(i+1,m)] for i in range(distributed_state.num_processes)])
'''
ext_separated_list_of_batches.append(split_into_batches) ext_separated_list_of_batches.append(split_into_batches)
''' '''
logger.info(f"\nDevice {distributed_state.device}: {ext_separated_list_of_batches}") logger.info(f"\nDevice {distributed_state.device}: {ext_separated_list_of_batches}")