Fine-tuning Llama 2 on Your Own Dataset | Train an LLM for Your Use Case with QLoRA on a Single GPU

Рет қаралды 121,675

Venelin Valkov

Күн бұрын

Пікірлер: 60

@venelin_valkov 10 ай бұрын

Full text tutorial (requires MLExpert Pro): www.mlexpert.io/prompt-engineering/fine-tuning-llama-2-on-custom-dataset

@kiranatmakuri3710 9 ай бұрын

Can you send me your email pls I have a question can’t ask in public

@williamfussell1956 8 ай бұрын

I keep having problems with the model.merge_and_unload()... It seems to be a bit different from the documentation on Hugging Face...is there something I am missing here? The error says that that 'LlamaForCausalLM' object has no attribute 'merge_and_unload'.... Any ideas?

@pablomeza5932 8 ай бұрын

@@williamfussell1956 Did you fix that?

@fl028 4 ай бұрын

@@williamfussell1956 merged_model = trained_model.merge_and_unload()

@christopherbader2222 9 ай бұрын

This is great. A version for question answering would be helpful too.

@echos01 10 ай бұрын

Excellent work! You are the hero!

@stawils 10 ай бұрын

Good stuff coming, thank you in advance ❤

@parisapouya6716 10 ай бұрын

Awesome work! Thanks a ton!

@HeywardLiu 10 ай бұрын

Awesome tutorial!

@vivekjyotibhowmik8008 10 ай бұрын

Can you provide the Google Collab notebook?

@NitroBrewbell 10 ай бұрын

very helpful. Thanks for the videos.

@VaibhavPatil-rx7pc 10 ай бұрын

Super excited

@krishchatterjee2819 9 ай бұрын

Excellent video! What changes in the input we need to make to use 8 bit quantization instead of 4 bit. Thanks.

@AbdulBasit-ff6tq 10 ай бұрын

Do you have or plan to make a tutorial for something like bellow? Tutorial for the plane text fine-tuning and then tuning that model to make it an instruct tuned one?

@GregMatoga 9 ай бұрын

Thank you for this! Is finetuning a good approach for a private/proprietary documentation Q&A?

@fabsync 2 ай бұрын

Fantastic video! It will be nice to see a full tutorial on how to do it with pdf locally...

@lyovazi8533 10 ай бұрын

very good video

@DawnWillTurn 10 ай бұрын

Any idea how can we deploy llama-2 on huggingface api? just like the falcon one, has some issue with the handler.

@williamgomezsantana 5 ай бұрын

Incredible video!! Thank you very much, I have a question: isn't it mandatory to put characters like EOS at the end of the summary? for the LLM to finish the instruction?

@jensonjoy83 10 ай бұрын

will you be able to add a tutorial for llama2-chat model

@experiment5762 9 ай бұрын

Great!! Do some videos regarding RLHF.

@techtraversal219 8 ай бұрын

Thanks for sharing, really helpful. Waiting for my Llama model access to follow it step by step. Can I use any other model in place of this?

@srushtiharyan2033 5 ай бұрын

Did you get the access? And how long did it take?

@ikurious 10 ай бұрын

Great video! Is there anyway to build my instruction dataset for instruct fine-tuning from classical text books?

@ikurious 10 ай бұрын

@@user-xt6tu3xt3t but then how to convert in question & answer format?

@mauriciososa9722 10 ай бұрын

@@ikurious the best way is manualyl by a human

@sasukeuchiha-ck4hy 10 ай бұрын

can you train the model on german data?

@tarunku9378 10 ай бұрын

I still don't get it i have my data locally , how should start finetuning it please tell

@chukypedro818 10 ай бұрын

Super🎉

@shopbc5553 7 ай бұрын

Do you have an idea how GPT4 is so good with its responses from its base model when I upload documents to it? Could it be the parameter. size only or do you think other technologies are what determine the quality difference?

@tillwill3232 Ай бұрын

parameter size and training data i guess? Also I dont think we know their exact network architecture since they didnt release their network publicly, can only access it via product

@GooBello-gr2ls 9 ай бұрын

can i download the finetuned model after finetuning? is it in format .bin or .safetensor or else? cuz im current trying to do finetuning on textgen, but having troubles. with dataset (format) i guess.

@lisab1360 9 ай бұрын

do you already know how you can download the finetuned model?

@williamfussell1956 8 ай бұрын

Hi there, I am just reading through the repo and Im pretty sure this is the answer...i just wanted to make sure... The actual input to the model is only from the [text] field, is that correct? As the [text] field contains the prompt, the conversation and the summary...

@user-xy5re6qh3d 7 ай бұрын

Hola, For me the validation log show No log with mistral instruct model. Please help anyone.

@danieladama8105 10 ай бұрын

🔥

@vitocorleon6753 10 ай бұрын

I need help please. I just want to be pointed in the right direction since I'm new to this and since I couldn't really find any proper guide to summarize the steps for what I want to accomplish. I want to integrate a LLama 2 70B chatbot into my website. I have no idea where to start. I looked into setting up the environment on one of my cloud servers(Has to be private). Now I'm looking into training/fine-tuneing the chat model using our data from our DBs(It's not clear for me here but I assume it involves two steps, first I have to have the data in a CSV format since it's easier for me, second I will need to format it in Alpaca or Openassistant formats). After that, the result should be a deployment-ready model ? Just bullet points I'd highly appreciate that.

@vitocorleon6753 10 ай бұрын

@nty3929 Oh :/ I’m still lost about this but thank you for your effort nevertheless!

@GregMatoga 9 ай бұрын

@nty3929 Yeah, bots are ruthless here and youtube is having none of it, even at that cost. Guess they expect to see more technical conversations elsewhere

@elysiryuu 5 ай бұрын

Thanks for the insight, is it possible to perform training locally, with 8 GB VRAM?

@stephenmartinez4883 3 ай бұрын

@karimbaig8573 8 ай бұрын

When you say you are tracking loss, what loss is that and how is that loss calculated for the task (summarization) at hand?

@anuranjankumar2904 6 ай бұрын

I have the same question. @karimbaig8573 were you able to figure out the answer?

@karimbaig8573 6 ай бұрын

Nope.

@tahahuraibb5833 10 ай бұрын

default_factory=lambda: ["q_proj", "v_proj"] Why did you not add this? Is it because HF does under the hood?

@venelin_valkov 10 ай бұрын

I totally forgot about the `target_modules`. I retrained and updated the notebook/tutorial with those. The results are better! Here's the list: lora_target_modules = [ "q_proj", "up_proj", "o_proj", "k_proj", "down_proj", "gate_proj", "v_proj", ] I composed it from here: github.com/huggingface/transformers/blob/f6301b9a13b8467d1f88a6f419d76aefa15bd9b8/src/transformers/models/llama/convert_llama_weights_to_hf.py#L144 Thank you!

@williamfussell1956 8 ай бұрын

Is there a good resource for understanding 'target modules' for different models? @@venelin_valkov

@cancheers 10 ай бұрын

should it be merged_model = trained_model.merge_and_unload()? cannot run, it is killed

@rone3243 10 ай бұрын

I have this problem as well😢

@kpratik41 10 ай бұрын

Were you able to solve this?

@fl028 4 ай бұрын

merged_model = trained_model.merge_and_unload()

@MecchaKakkoi 5 ай бұрын

This looks like a great notebook, however, I always get a "CUDA out of memory" error when it executes the SFTTrainer function. It's fine up until then according to nvidia-smi but then memory just instantly maxes out. Does anyone know a way around this?

@rishabjain9275 5 ай бұрын

try reducing the sequence length

@fl028 4 ай бұрын

I reduced per_device_train_batch_size=1,

@okopyl 9 ай бұрын

Why do you use that kind of prompt for the training like `### Instruction`? When in fact Llama 2 prompts are like `[INST] `...

@g1rlss1mp 9 ай бұрын

I think it's a LLaMA2-CHAT prompt. The base model was not finetuned.

@skahler 9 ай бұрын

omg @ 15:06 😂😂😂

@JeeneyAI 5 ай бұрын

ALL of these tutorials require more dependencies. Can't somebody post how to do this in pycharm with your own GPU? I can't make any of the tutorials I've found work and it's just an endless troubleshooting process as to why everything is different in all of them