hoelee/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2025-08-08 13:19:54 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add ability to choose using weighted loss or not

Browse Source
This commit is contained in:
Shondoit
2023-01-12 16:29:00 +01:00
parent bc50936745
commit edb10092de
4 changed files with 32 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
modules/textual_inversion/textual_inversion.py
View File

@@ -351,7 +351,7 @@ def validate_train_inputs(model_name, learn_rate, batch_size, gradient_step, dat
assert log_directory, "Log directory is empty"
def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_step, data_root, log_directory, training_width, training_height, varsize, steps, clip_grad_mode, clip_grad_value, shuffle_tags, tag_drop_out, latent_sampling_method, create_image_every, save_embedding_every, template_filename, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_step, data_root, log_directory, training_width, training_height, varsize, steps, clip_grad_mode, clip_grad_value, shuffle_tags, tag_drop_out, latent_sampling_method, use_weight, create_image_every, save_embedding_every, template_filename, save_image_with_stored_embedding, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height):
save_embedding_every = save_embedding_every or 0
create_image_every = create_image_every or 0
template_file = textual_inversion_templates.get(template_filename, None)
@@ -410,7 +410,7 @@ def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_st
pin_memory = shared.opts.pin_memory
ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method, varsize=varsize)
ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, cond_model=shared.sd_model.cond_stage_model, device=devices.device, template_file=template_file, batch_size=batch_size, gradient_step=gradient_step, shuffle_tags=shuffle_tags, tag_drop_out=tag_drop_out, latent_sampling_method=latent_sampling_method, varsize=varsize, use_weight=use_weight)
if shared.opts.save_training_settings_to_txt:
save_settings_to_file(log_directory, {**dict(model_name=checkpoint.model_name, model_hash=checkpoint.shorthash, num_of_dataset_images=len(ds), num_vectors_per_token=len(embedding.vec)), **locals()})
@@ -480,7 +480,8 @@ def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_st
with devices.autocast():
x = batch.latent_sample.to(devices.device, non_blocking=pin_memory)
w = batch.weight.to(devices.device, non_blocking=pin_memory)
if use_weight:
w = batch.weight.to(devices.device, non_blocking=pin_memory)
c = shared.sd_model.cond_stage_model(batch.cond_text)
if is_training_inpainting_model:
@@ -491,7 +492,11 @@ def train_embedding(id_task, embedding_name, learn_rate, batch_size, gradient_st
else:
cond = c
loss = shared.sd_model.weighted_forward(x, cond, w)[0] / gradient_step
if use_weight:
loss = shared.sd_model.weighted_forward(x, cond, w)[0] / gradient_step
del w
else:
loss = shared.sd_model.forward(x, cond)[0] / gradient_step
del x
_loss_step += loss.item()