Activity
Mon
Wed
Fri
Sun
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
What is this?
Less
More
Memberships
Data Alchemy
Public • 23.6k • Free
7 contributions to Data Alchemy
19d ago in
LLM Get Exact Response
Hi everyone, I am working on getting an exact response from the Llama3.1:8b-instruct. I have set the following parameters: - temperature = 0 - top_k = 1 - seed = 0 but the results vary. Out of 10 requests, I got responses like ~3, ~3, ~3, and ~1, which are quite similar. Any ideas, to get exactly same responses Thanks 😊
0
1
New comment 19d ago
Aug 8 in
Robust LLM output
Hi everyone, I am working on a project using LLaMA 3:8B or LLaMA 3.1:B with Ollama. The issue I'm facing is ensuring the output is robust and follows the exact format specified. For example, when I provide a prompt that should force the following output: ```json { "name": "Ali", "age": 20, "major": "CS" } ``` The actual output sometimes looks like this due to variations in the input: ```json { "name": "Ali", "city": "xxx" } ``` as shown it not return certain expected fields (if not exist its not return '' ) and it add field not asked in prompt output format. any ideas thanks
6
4
New comment Aug 10
0 likes • Aug 9
thanks for replay I am use ollama or direct using requests response = ollama.chat( model="llama3.1", options=options, # Pass the options object here messages= [ {"role": "system","content": prompt_c}, {"role": "user", "content": assistance_user_1}, {"role": "assistant", "content": json.dumps(assistance_generate_1)}, {"role": "user","content": user_text}, ], format="json",}
0 likes • Aug 10
thanks so much I will test this 😁
Sep '23 in
Anyone try GitHub Copilot for learning of Coding?
I am considering subscribing to GitHub Copilot and GitHub Copilot Lab after watching Dave’s video to enhance my coding skills in Machine Learning and Deep Learning. However, I would also like to hear from anyone who has experience, good or bad, with these tools before I make a decision. Thanks, Joe
8
9
New comment Sep '23
4 likes • Sep '23
CoPilot integrates tightly with the VS Code environment, giving me the sensation of wanting to craft code, even when I'm uncertain about the exact syntax. However, GPT offers a distinct advantage in its flexibility. Not only do I use it for coding and debugging, but it also aids in the writing process, making it smoother and more intuitive. With GPT, I'm empowered to center my attention on the underlying problem, freeing me from the constraints of code intricacies and letting the writing flow naturally.
Sep '23 in
MATHS!!
Extremely recently I've started my "data science" journey and I've found maths is heavily involved. What key math concepts should beginners focus on, and how can I improve my math skills?
11
17
New comment 19d ago
7 likes • Sep '23
Linear Algebra, Calculus, statistics and probability
4 likes • Sep '23
Just the basic is enough What I mean For linear Algebra If you know really what the vectors and metrics and how make math operations on them this enough Calculus If you know what the Integration mean ( areas under curves or base function for some differential function) Differentiation-> slope or gradient for function Probability and statistics Basic knowledge on each Krish amazing in the journey from basics to deployment topics! https://m.youtube.com/watch?v=4x30Fziru6M&pp=ygUZa3Jpc2ggbmFpayBtYXRocyBwbGF5bGlzdA%3D%3D
Sep '23 in
Which LLM do you use most often?
I've tried a variety of open-source LLMs, but I've found that GPT-3.5 gives the best responses, and sometimes only GPT-4 performs better. What about you?
12
18
New comment Oct '23
1 like • Sep '23
@Miguel Tajada Ferrer GPT-4 in most cases 😅
1-7 of 7
Active 11d ago
Joined Sep 15, 2023
powered by