#data-assistants-with-gemma

1 messages · Page 1 of 1 (latest)

sick tangle
#

when i input "gemma_lm = keras_nlp.models.GemmaCausalLM.from_preset("gemma_2b_en")" in kaggle, output error "AttributeError: module 'keras_nlp.models' has no attribute 'GemmaCausalLM", why?

charred wharf
#

Hello everyone. I can't import the Gemma Model to start working. When I try with Keras, it gives me the same error as in @sick tangle

earnest hinge
#

Do you guys all already have teams for this competition ?

fathom crag
strange spindle
#

Is anyone else uncertain about the expected output for this? I know they include the guidelines and rubric but it still seems very vague

turbid stone
fair river
#

Hey everyone!

I'm thrilled to present my latest notebook: "Gemma, LLM, Kaggle Solution Writeup Agent"!

Ever wondered how top performers crack the code in Kaggle competitions? Join me as we'll unlock the secrets behind winning solutions, powered by Gemma, Google's Open Model.
We'll unravel the strategies, techniques, and secrets behind those winning submissions. With Gemma we'll dive deep into the heart of data science excellence.

Throughout the notebook, we'll:

🔍 Deep dive into Kaggle solution write-ups, dissecting the approaches of top performers.
🎓 Explore innovative methods for data generation, few-shot prompting, and fine-tuning LLMs to boost model performance.
💡 Break down complex concepts with real-life examples, making them accessible to everyone interested in data science.
Discover how to upload your fine-tuned models to the Kaggle Model Hub, opening up opportunities for collaboration and knowledge-sharing within the community.

Check out the notebook right here: https://www.kaggle.com/code/ianakoto/gemma-llm-kaggle-solution-writeup-agent

Gemma, LLM, Kaggle Solution Writeup Agent

Explore and run machine learning code with Kaggle Notebooks | Using data from multiple data sources

modest oracle
#

Hi everyone,

I just wanted to ask if someone can clarify this part of the rules to me "We will grade the most recent submission from your team (per competition, not per category)."

I am not sure if this means that we can only have one submission per competition? or if participants can participate in each category, for instance, if I can submit a project per category, meaning 5 projects?

Just in case, English is my second language 🙂

Please let me know. Thank you in advance.

turbid stone
#

The competition allow you to participate in one category only; but you can submit multiple times to any category you want. only the last submission you make and the selected category for it will be considered. Let’s assume you submit in notebook 1 to category 1 then after a week you submitted notebook 2 to category 2 . Only category 2 and notebook 2 will be considered in judging.

modest oracle
#

Thank you very much for your answer @turbid stone . Have a nice day

modest oracle
#

A question, for the competition do I am force to use the code that is in kaggle, I mean this code presented in this kaggle link (https://www.kaggle.com/models/google/gemma)

import keras
import keras_nlp
import numpy as np
gemma_lm = keras_nlp.models.GemmaCausalLM.from_preset("gemma_instruct_2b_en")
gemma_lm.generate("Keras is a", max_length=30)

Generate with batched prompts.

gemma_lm.generate(["Keras is a", "I want to say"], max_length=30)
Compile the generate() function with a custom sampler.

Or can I use the code presented in hugging face that uses gemma model as well with a different code?

Thank you vey much in advance for your answers.

Gemma

Gemma is a family of lightweight, open models built from the research and technology that Google used to create the Gemini models.

modest oracle
#

Hi everyone,

Just to confirm. This gemma model does not have memory like ChatGPT, right? or is there a way for this to have memory?

Thank you in advance for your answers

bold lion
#

Yes, LLMs don't have long term memory. You can add one by maintaining a list which store history of both user and assistant.

bold lion
craggy owl
ivory summit
#

Hi everyone,

Can I update the notebook after submitting? I noticed a lot of typos and the build when saving the notebook produced a slightly different output which makes an explanation I made inaccurate.

foggy pewter
ivory summit
foggy pewter
#

well now people know why you're asking and they can reply , ofc i dunno ;D

modest oracle
#

Hi there!

I am new to Kaggle and I am not familiar yet. I wonder why here in the code tab I can see other notebooks except mine?

I looked for my notebook and I did not find the my notebook. Could you let me know what happened?

Thank you in advance

hazy bane
#

The code tab is to see noteooks that have been publicly shared with everyone

inland summit
#

Hello everyone, I've been trying to get feedback from Kaggle staff in order to better understand the final leaderboard and awards. Here my pending inquiry: https://www.kaggle.com/competitions/data-assistants-with-gemma/discussion/499090#2787513.

In a nutshell, I'd like to know how the winning solution of the Summarization category scored in the Eligibility Criteria under the "Open", since the fine-tuned model and its training methodology do not appear to be open.

The forum seems to be no longer active; could someone kindly reach out to me? Sorry for the tag, @hazy bane , but you are the only Kaggle stuff I can see in this channel.

hazy bane
#

@inland summit Sorry I don't work on competitions directly so I can't help you. The forums are the best way to make sure your message is seen.

inland summit
#

thanks for you kind reply, unfortunately I'm getting no answers from days..perhaps is there another Kaggle staff here I can contact directly?

hazy bane
#

Not in discord I'm afraid. I'm pretty sure the staff on the competition have seen your message but they might not respond right away.