Comments (9)
Hummm... might be interesting co write a validator for things like this to raise an error when the syntax does not comply with what sd-scripts expect... or maybe just to remove the offending characters... I will think about it... should not be too difficult...
from kohya_ss.
Okay, but the parameter for betas looks like this, so I can't get around the comma: (0.9, 0.999) https://github.com/konstmish/prodigy So if I need the comma, what's the solution? I do not use toml.
"betas=(0.9,0.99)"
You can't have a space in the brackets.
from kohya_ss.
Make sure Thera are no comma between each group of parameters. Can you share the json config that cause the issue?
from kohya_ss.
Here is my config, I deleted personal paths. There's a comma in betas parameter. This was not a problem in previous versions:
{
"LoRA_type": "Standard",
"LyCORIS_preset": "full",
"adaptive_noise_scale": 0,
"additional_parameters": "",
"block_alphas": "",
"block_dims": "",
"block_lr_zero_threshold": "",
"bucket_no_upscale": true,
"bucket_reso_steps": 64,
"bypass_mode": false,
"cache_latents": true,
"cache_latents_to_disk": false,
"caption_dropout_every_n_epochs": 0,
"caption_dropout_rate": 0,
"caption_extension": ".txt",
"clip_skip": 1,
"color_aug": false,
"constrain": 0,
"conv_alpha": 1,
"conv_block_alphas": "",
"conv_block_dims": "",
"conv_dim": 1,
"dataset_config": "",
"debiased_estimation_loss": false,
"decompose_both": false,
"dim_from_weights": false,
"dora_wd": false,
"down_lr_weight": "",
"enable_bucket": false,
"epoch": 10,
"extra_accelerate_launch_args": "",
"factor": -1,
"flip_aug": false,
"fp8_base": false,
"full_bf16": true,
"full_fp16": false,
"gpu_ids": "",
"gradient_accumulation_steps": 1,
"gradient_checkpointing": false,
"huber_c": 0.1,
"huber_schedule": "snr",
"ip_noise_gamma": 0,
"ip_noise_gamma_random_strength": false,
"keep_tokens": 0,
"learning_rate": 1,
"log_tracker_config": "",
"log_tracker_name": "",
"lora_network_weights": "",
"loss_type": "l2",
"lr_scheduler": "cosine",
"lr_scheduler_args": "",
"lr_scheduler_num_cycles": "",
"lr_scheduler_power": "",
"lr_warmup": 0,
"main_process_port": 0,
"masked_loss": false,
"max_bucket_reso": 2048,
"max_data_loader_n_workers": "0",
"max_grad_norm": 1,
"max_resolution": "512,512",
"max_timestep": 1000,
"max_token_length": "225",
"max_train_epochs": "",
"max_train_steps": "",
"mem_eff_attn": false,
"mid_lr_weight": "",
"min_bucket_reso": 512,
"min_snr_gamma": 5,
"min_timestep": 0,
"mixed_precision": "bf16",
"model_list": "custom",
"module_dropout": 0,
"multi_gpu": false,
"multires_noise_discount": 0,
"multires_noise_iterations": 0,
"network_alpha": 64,
"network_dim": 64,
"network_dropout": 0,
"noise_offset": 0,
"noise_offset_random_strength": false,
"noise_offset_type": "Original",
"num_cpu_threads_per_process": 2,
"num_machines": 1,
"num_processes": 1,
"optimizer": "Prodigy",
"optimizer_args": "use_bias_correction=True weight_decay=0.5 d_coef=0.65 \"betas=(0.9, 0.901)\"",
"output_name": "xxxxxx",
"persistent_data_loader_workers": false,
"pretrained_model_name_or_path": "M:/!SDModels/!Edzeni/realisticVisionV51_v51VAE_full.safetensors",
"prior_loss_weight": 1,
"random_crop": false,
"rank_dropout": 0,
"rank_dropout_scale": false,
"rescaled": false,
"resume": "",
"sample_every_n_epochs": 1,
"sample_every_n_steps": 0,
"sample_prompts": "a xxxxxx woman in restaurant, upper body, red hair, realistic photo, many people in background, RAW photo, 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3 --n lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry",
"sample_sampler": "euler_a",
"save_as_bool": false,
"save_every_n_epochs": 1,
"save_every_n_steps": 0,
"save_last_n_steps": 0,
"save_last_n_steps_state": 0,
"save_model_as": "safetensors",
"save_precision": "bf16",
"save_state": false,
"save_state_on_train_end": false,
"scale_v_pred_loss_like_noise_pred": false,
"scale_weight_norms": 1,
"sdxl": false,
"sdxl_cache_text_encoder_outputs": false,
"sdxl_no_half_vae": true,
"seed": "1217930809",
"shuffle_caption": false,
"stop_text_encoder_training": 0,
"text_encoder_lr": 1,
"train_batch_size": 1,
"train_norm": false,
"train_on_input": true,
"training_comment": "",
"unet_lr": 1,
"unit": 1,
"up_lr_weight": "",
"use_cp": false,
"use_scalar": false,
"use_tucker": false,
"use_wandb": false,
"v2": false,
"v_parameterization": false,
"v_pred_like_loss": 0,
"vae": "",
"vae_batch_size": 0,
"wandb_api_key": "",
"wandb_run_name": "",
"weighted_captions": false,
"xformers": "xformers"
}
from kohya_ss.
Yeah, this new version if using the toml and it changes the rule for parameters... don't put " or , in there... This should fix the issue...
from kohya_ss.
Okay, but the parameter for betas looks like this, so I can't get around the comma: (0.9, 0.999)
https://github.com/konstmish/prodigy
So if I need the comma, what's the solution? I do not use toml.
from kohya_ss.
The dev
branch code will now validate the lr scheduler and optimizer arguments and prevent starting the training if they do not comply with the needed format.
The empty fields for those two optional arguments show what format arguments should have.
from kohya_ss.
The
dev
branch code will now validate the arguments and prevent starting the training if they do not comply with the needed format.
This raises an interesting possibility. These systems have lots of arguments that can be leveraged for all sorts of purposes. This would probably be a big as, but would it be possible have a list and the correct formating. Like in a separate tab. This would help with user's troubleshooting efforts. Yes, the concept of having to documentation is always a big ask; but in general users should always know the correct means of interfacing with the system within the system.
I assume there is only limited and known amount and type of arguments which can be or are passed through?
from kohya_ss.
I wish kohya did. It would really be a document he should maintain.
from kohya_ss.
Related Issues (20)
- avr_loss=nan
- Debian 12 is the new for this HOT 3
- logging wandb is now worst
- Pytorch version inconsistency on google colab notebook
- Bad file descriptor in git clone
- BLIP2 captioning doesn't see .webp files (with patch) HOT 1
- Relationship between toml and json config files HOT 3
- I get an error message that a url: /api/models/stabilityai/stable-diffusion-xl-base-1.0 is missing when i start lora training local on windows without internet. HOT 1
- TypeError: environment can only contain strings
- 49 sec/it on 4090 HOT 3
- ValueError: Key backend: 'module://matplotlib_inline.backend_inline' is not a valid value for backend HOT 1
- assert abs_path is not None, f"no image / 画像がありません: {image_key}" HOT 1
- Didn't change anything, Lora training spits out non-zero exit status 1 json error? HOT 1
- ValueError: Too many values to unwrap (expected 2)...how to fix
- {Help} Network Rank Slider and No Half VAE checkbox is not showing up in my Kohya_ss after Installation. HOT 1
- AttributeError: 'NoneType' object has no attribute 'dtype' HOT 2
- Additional Parameters "Scheduler args" NOT Saved in Toml file AND NOT passed HOT 8
- No json files on training HOT 4
- WD14 Tagger Not Recognizing CUDA After Update? HOT 12
- sdxl training HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from kohya_ss.