good agentic datasets
Collection
14 items
•
Updated
•
1
John
juyongjiang
AI & ML interests
LLMs for Code Generation
Recent Activity
upvoted
a
collection
about 16 hours ago
good agentic datasets
liked
a model
1 day ago
inclusionAI/LLaDA2.1-mini
published
a dataset
6 months ago
juyongjiang/verified-code-data
Organizations
upvoted
a
collection
about 16 hours ago
liked
a
model
1 day ago
published
a
dataset
6 months ago
updated
a
dataset
6 months ago
upvoted
a
paper
10 months ago
liked
a
Space
11 months ago
authored
5
papers
about 1 year ago
HyperCLOVA X Technical Report
Paper
•
2404.01954
•
Published
•
25
Shortcut-connected Expert Parallelism for Accelerating Mixture-of-Experts
Paper
•
2404.05019
•
Published
A Survey on Large Language Models for Code Generation
Paper
•
2406.00515
•
Published
A Survey on Mixture of Experts
Paper
•
2407.06204
•
Published
KaSA: Knowledge-Aware Singular-Value Adaptation of Large Language Models
Paper
•
2412.06071
•
Published
•
9
liked
a
model
about 1 year ago
updated
a
model
about 1 year ago
upvoted
a
paper
about 1 year ago
upvoted
an
article
over 1 year ago
Article
Introducing the Open Ko-LLM Leaderboard: Leading the Korean LLM Evaluation Ecosystem
- +1
•
5
liked
a
model
over 1 year ago
reacted to
chansung's
post with 🚀➕🔥👍
over 1 year ago
Post
4030
🦙🦙 LLaMA Duo project update
Last time, I gave a brief introduction about LLaMA Duo project with @sayakpaul . It is a simple toolset to aligning sLLM with service LLM with coverage dataset 👉🏻 (https://huggingface.co/posts/chansung/708646454991943).
- coverage dataset is what we believe to be the most important/desired (instruction, response) pairs. In system thinking, each instruction could be an analogy of a function from traditional programming. We make unit tests and measure the coverage % for all the features/functions. Similarly, we need to ensure if our fine-tuned model could handle what % of given instructions from coverage dataset satisfactory (hence coverage dataset).
We have tested it with "Coding" category of data from HuggingFaceH4/no_robots dataset. It has about 300 SFT training data points under Coding category. After fine-tuning Gemma 7B model on that, the result was very poor. LLaMA Duo's evaluation tool gave < 20% of metrics in similarity and preciseness on the test split.
So, we used LLaMA Duo's synthetic data generation tool to generate 60k data points that looks similar to the original dataset. We first created ~10k synthetic data points, then created 50k more based on the synthetic dataset itself.
After fine-tuning Gemma 7B on the 60k synthetic dataset, the evaluation result went up to 80~90% high. Also, when testing out the model in UI, it tends to give good responses.
It is a good showcase to transition from service LLM to sLLM or having a backup sLLM for service LLM failure scenarios. I am going to expand this experiments on all categories of no_robots dataset. It will roughly generate > 100k data points.
Here are some links:
- LLaMA Duo project repo: https://github.com/deep-diver/llamaduo
- 60k Coding synthetic dataset: chansung/merged_ds_coding
- Fine-tuned Gemma 7B model: chansung/coding_llamaduo_60k_v0.2
Last time, I gave a brief introduction about LLaMA Duo project with @sayakpaul . It is a simple toolset to aligning sLLM with service LLM with coverage dataset 👉🏻 (https://huggingface.co/posts/chansung/708646454991943).
- coverage dataset is what we believe to be the most important/desired (instruction, response) pairs. In system thinking, each instruction could be an analogy of a function from traditional programming. We make unit tests and measure the coverage % for all the features/functions. Similarly, we need to ensure if our fine-tuned model could handle what % of given instructions from coverage dataset satisfactory (hence coverage dataset).
We have tested it with "Coding" category of data from HuggingFaceH4/no_robots dataset. It has about 300 SFT training data points under Coding category. After fine-tuning Gemma 7B model on that, the result was very poor. LLaMA Duo's evaluation tool gave < 20% of metrics in similarity and preciseness on the test split.
So, we used LLaMA Duo's synthetic data generation tool to generate 60k data points that looks similar to the original dataset. We first created ~10k synthetic data points, then created 50k more based on the synthetic dataset itself.
After fine-tuning Gemma 7B on the 60k synthetic dataset, the evaluation result went up to 80~90% high. Also, when testing out the model in UI, it tends to give good responses.
It is a good showcase to transition from service LLM to sLLM or having a backup sLLM for service LLM failure scenarios. I am going to expand this experiments on all categories of no_robots dataset. It will roughly generate > 100k data points.
Here are some links:
- LLaMA Duo project repo: https://github.com/deep-diver/llamaduo
- 60k Coding synthetic dataset: chansung/merged_ds_coding
- Fine-tuned Gemma 7B model: chansung/coding_llamaduo_60k_v0.2