Hey Yuval, this is super helpful. Thank you so much! Trying right now. Also no worries I'm a tech bro and you're not half stupid lol, you're helping a bunch of people here so 🙏
at 3:54 you chose "public". Is this required? I did private, but getting errors when using the model and it's generating empty.zip and no output. Wondering if this could be my issue?
Hello, YUVAL!!!!! very good! I have a doubt. Is it possible, for example, to have Rayban sunglasses in the same folder for you to be using or do I have to do different training? How can I do this? Thanks.!
so it will upload the LOrA that you trained to huggingFace hub, so afterwards, when you use the other job to create your image, it will fetch the LoRa and will use it
Great video! Pls tell me you know how we can rename our layers?? This is cray not seeing an option to do so and in my Pro version, such a basic thing to stay organized
I followed your process on replicate to try and make LoRA models but when I try to use them there (just as you did) I get an error. It tells me that my file does not appear to have a file named pytorch_lora_weights.bin.
Hey Yuval, Great Video mate! Do I have to pay a Hugging Face account to run it? My traning seems to be failing. I asked chat gpt and it says that that while I'm passing a Hugging Face token, the error indicates that it doesn¿t have the necessary permissions to access the model. Thanks for the feedback man!
@luisfernandocastanedagonza5 did you figure it out? I followed the steps, created the model successfully (I think, after a few failed attempts). Now when I try to create a prompt it runs and the output is "empty.zip". No errors, just no files. I didn't see in Yuval's video a linking of a Flux subscription to Replicate, so maybe I missed something there?
how to generate images with more people and these random people? After I created my Lora, when I go to create an image at a party, for example, all the people around have my face, how do I solve it?
Yes becuse that's the request from the AI took kit training library, I am not a big tech person but seems like it's the best solution especially if you want to call it later using the API. Do you have a guide for generating images with replicated stored without hugging face?
@@aimakerslab yes by default you can store the fine tuned models within replicate! You can also have them as private, while on HF they need to be public.
Can you make a video on how to do an accurate product model training? Something like the Coca Cola can but do it for a lesser known brand with a bunch of text and maybe a slightly different type of packaging. I really want to see consistency in Products while generating using Flux
Got it, automate with make and airtable. I wonder if you can do an video or show example of the training images required for products, and the way you have described the images for the AI. Thanks.
You use generative ai and your TH-cam thumbnails look just like every else who read the TH-cam thumbnail guide. I gave your video a shot and it’s good but I usually skip those cause they are like a TH-cam thirst trap, but less creative than a celebrity nip slip. I think you’d do slightly better with a more creative thumbnail. Content is already good, try it, A/B test it. Be not like the other trillion TH-cam thumbnails.
I was SO pumped for this, watched again and followed all just as explained, then when running I get this 'Output Preview JSON Training failed. Command '['python', 'run.py', 'config/replicate.yml']' returned non-zero exit status 1.
You didn't set learning rate, steps, lora linear and lora alpha to name some. There are better guides out there that gives you information about those presets.
Yes, it's not free. They are basically renting out a poweful graphics card to you which is great so you don't have to spend thousands of dollars in buying the hardware yourself.
Fan-darn-tastic, thank you for a great tutorial! BTW, loved the Yuval in then lavender fields! 👏👏👍👍
Hey Yuval, this is super helpful. Thank you so much! Trying right now.
Also no worries I'm a tech bro and you're not half stupid lol, you're helping a bunch of people here so 🙏
really lovin your content. Great tutorials.
Amazing! 🎉
at 3:54 you chose "public". Is this required? I did private, but getting errors when using the model and it's generating empty.zip and no output. Wondering if this could be my issue?
You can use private, I have successfully.
love you Yuval
Is there any way to do it for free?
How to use the Lora with the new model flux-1.1-pro-ultra?
Hello, YUVAL!!!!! very good!
I have a doubt. Is it possible, for example, to have Rayban sunglasses in the same folder for you to be using or do I have to do different training? How can I do this?
Thanks.!
I tried, but "Training failed. Command '['python', 'run.py', 'config/replicate.yml']' returned non-zero exit status 1.
check the resolution of the images you input, if it's bigger than 1024, you'd have to specify where it says resolution,
Had the same message, did the process 10 times, could resolve it by renaming each of my pictures, I had to use "_" instead of "-" now it's working.
Try going to the black-forest-labs hugging face page and accepting the conditions of use. This solved the issue for me
@@kylewalker4641 Thanks, works now for me
Why do you need a repo ID and tokens from Hugging Face? What are their purposes?
so it will upload the LOrA that you trained to huggingFace hub, so afterwards, when you use the other job to create your image, it will fetch the LoRa and will use it
It's not necessary, you can download it after the training is done from Replicate.
Love it!! Amazing !! is that the coke can using the same training model? thanks Yuval!
Do you have to name each photos before you zip in a certain format
Yes. Captioning is part of it.
Great video! Pls tell me you know how we can rename our layers?? This is cray not seeing an option to do so and in my Pro version, such a basic thing to stay organized
כל הכבוד אחי :) אני גיליתי את זה לפני כמה חודשים וזה משנה משחק, אם אתה רוצה אני אשמח לשתף איתך פעולה
I followed your process on replicate to try and make LoRA models but when I try to use them there (just as you did) I get an error. It tells me that my file does not appear to have a file named pytorch_lora_weights.bin.
Me too. Did you manage to solve it?
Hi Yuval!! Thanks for the explanation.. can we use the result trainning for commercial ? hope you reply with answer,, thanks before.. cheers
read about pricing for the image was generating in any model in replicate document
Is it paid?
I want to create two consistent characters appearing in one image, please help me
Hey Yuval, Great Video mate! Do I have to pay a Hugging Face account to run it? My traning seems to be failing. I asked chat gpt and it says that that while I'm passing a Hugging Face token, the error indicates that it doesn¿t have the necessary permissions to access the model. Thanks for the feedback man!
@luisfernandocastanedagonza5 did you figure it out? I followed the steps, created the model successfully (I think, after a few failed attempts). Now when I try to create a prompt it runs and the output is "empty.zip". No errors, just no files. I didn't see in Yuval's video a linking of a Flux subscription to Replicate, so maybe I missed something there?
You don't need Huggingface at all. The model is saved on Replicate.
how to generate images with more people and these random people? After I created my Lora, when I go to create an image at a party, for example, all the people around have my face, how do I solve it?
Is there a reason you are using hugging face instead of storing your model on replicate?
Yes becuse that's the request from the AI took kit training library,
I am not a big tech person but seems like it's the best solution especially if you want to call it later using the API.
Do you have a guide for generating images with replicated stored without hugging face?
@@aimakerslab yes by default you can store the fine tuned models within replicate! You can also have them as private, while on HF they need to be public.
@@HelloFromNick I'll look into it
why we need huggingface what it use?
Can you download the model from replicate to use locally after training it?
Yes, that's what I have been doing all the time.
Can you make a video on how to do an accurate product model training? Something like the Coca Cola can but do it for a lesser known brand with a bunch of text and maybe a slightly different type of packaging.
I really want to see consistency in Products while generating using Flux
Where's the video about the coke can? I can't see it on your channel?
Got it, automate with make and airtable. I wonder if you can do an video or show example of the training images required for products, and the way you have described the images for the AI. Thanks.
You use generative ai and your TH-cam thumbnails look just like every else who read the TH-cam thumbnail guide. I gave your video a shot and it’s good but I usually skip those cause they are like a TH-cam thirst trap, but less creative than a celebrity nip slip. I think you’d do slightly better with a more creative thumbnail. Content is already good, try it, A/B test it. Be not like the other trillion TH-cam thumbnails.
I was SO pumped for this, watched again and followed all just as explained, then when running I get this 'Output
Preview
JSON
Training failed.
Command '['python', 'run.py', 'config/replicate.yml']' returned non-zero exit status 1.
@@thewebstylist what i often do is taking a screenshot, paste it to claude or chatgpt and ask it
@@thewebstylist id guess its related to the output of cladue so try to reiterate the system prompt
so you cant download your lora?
Yes you can, huggingface isn't necessary.
Do you use the paid or free version of replicate?
@@naturallydope247 paid
Is this totally free or not ???
No
can we do this thing is free please tell me
Doing it on your own computer locally but then you need a different guide. 6Gb VRAM, expect your computer to be busy for 3 straight days.
Can you download the safetensors file after?
Yes
You didn't set learning rate, steps, lora linear and lora alpha to name some.
There are better guides out there that gives you information about those presets.
But it is paid. How can we use it for free?
free replicate???
Doing it on your own computer locally but then you need a different guide. 6Gb VRAM, expect your computer to be busy for 3 straight days.
Not free
Merci bro
good
Completely unrelated…. But where did you buy that shirt? lol
@@playactionstudios i go with Uniqlo.
Bought like 5 of these. Good eye!
Would you have a $10 free invitation to replicate, please?
Replicate is asking money
Yes, it's not free. They are basically renting out a poweful graphics card to you which is great so you don't have to spend thousands of dollars in buying the hardware yourself.
It's actually Low Rank *ADAPTATION*
@@erfanasgari21 thanks!!