Create Synthetic Data to Train DeepSeek R1! (Reasoning Steps)

Поділитися
Вставка
  • Опубліковано 5 лют 2025

КОМЕНТАРІ • 13

  • @rousabout7578
    @rousabout7578 7 днів тому +2

    Thanks Mate! While everyone else is focused on the drama, Mervin Praison is leveling up!

  • @climateireland7546
    @climateireland7546 8 днів тому +5

    Super video, would love to see a part two about how to use the training dataset to train a model of choice.
    Also, it would be great to go into a little more detail as to how you can use custom docs and CSVs to build out the training reasoning questions…. But maybe I just need to watch it again 😅

  • @SurajLambor-mq2mf
    @SurajLambor-mq2mf 6 днів тому +5

    I tried this method, but I got an error when running the agent. It creates the question and answer but does not push them into the CSV and I receive the following error:Traceback (most recent call last):
    File "D:\Data\app.py", line 146, in
    agents.start()
    File "D:\Data\myenv\lib\site-packages\praisonaiagents\agents\agents.py", line 817, in start
    self.run_all_tasks()
    File "D:\Data\myenv\lib\site-packages\praisonaiagents\agents\agents.py", line 769, in run_all_tasks
    for task_id in process.workflow():
    File "D:\Data\myenv\lib\site-packages\praisonaiagents\process\process.py", line 888, in workflow
    self._create_loop_subtasks(current_task)
    AttributeError: 'Process' object has no attribute '_create_loop_subtasks'

  • @danielcamposramos9943
    @danielcamposramos9943 7 днів тому

    Thank you so much master for sharing this! It was the missing step for me, now no complaints anymore for the lack of a UI. Keep the wonderful work master!

  • @chidi21
    @chidi21 8 днів тому

    Very cool Marvin 👍, this video came in just in time. Really useful stuff here!

  • @gbengaomoyeni4
    @gbengaomoyeni4 8 днів тому +1

    Mervin, you are super genius! I appreciate your work. Please, can you amke a tutorial on parsing a pdf document containing text and images and it should fetch each user with the information about them in either database or csv or saved the parsed data in json

  • @spicer41282
    @spicer41282 8 днів тому

    You read my mind wish! Thank you!

  • @AndreGenesini
    @AndreGenesini 8 днів тому +4

    excellent Marvin, as always! Two questions please: Deeopseek R1 also used some reinforcement if it took the right decision or solution, how to mimic that? Also how to generate questions and reasoning from a rag material source, like pdfs that teach some line of tought like psychoanalisys and clinical evaluation of psyshoanalitical cases?

  • @MotivateGPT
    @MotivateGPT 8 днів тому +1

    how to train deep seek mode;// please provide video

  • @ravirajguru7044
    @ravirajguru7044 7 днів тому

    Great and Thanks . Already viewed your recommend. I try to implement every video. but as I use windows 10. some work but for other n number of issues popup. Please do something about it.

  • @ngana8755
    @ngana8755 4 дні тому

    The steps are confusing. It was hard to follow along. Also, what was the purpose of creating a DeepSeek base model with reasoning capabilities when DeepSeek released R1 with reasoning capabilities? It sounds like reinventing the wheel.

  • @sd5853
    @sd5853 8 днів тому +1

    powerful