Skip to content

Instantly share code, notes, and snippets.

View vishalsodani's full-sized avatar
🏠
Working from home

Vishal Sodani vishalsodani

🏠
Working from home
View GitHub Profile
@vishalsodani
vishalsodani / kutch.md
Last active November 23, 2023 07:07
kutch
@vishalsodani
vishalsodani / mami2023.md
Created October 25, 2023 06:44
mami 2023
  1. Animal Kingdom (France)
  2. Beyond Utopia (South Korea)
  3. Blackbird Blackbird Blueberry (Georgia)
  4. Bye Bye Tiberias (Palestine)
  5. Dormitory (Turkey)
  6. Fremont (US)
  7. If I Could Hibernate (Mongolia)
  8. Inshallah a Boy (Jordan)
  9. Kokomo City (US)
  10. Sweet Dreams (The Netherlands)
@vishalsodani
vishalsodani / leakage complaints.md
Created October 20, 2023 05:02
leakage complaints

leakage complaints kindly write a complaint letter to Deputy Registrar of Co-operative Societies, against your society, for not taking any action for leakage in your flat from Chairman's residence.

Under Bye-law No. 174(D)(iv) same complaint in same wordings, address it to Assistant Municipal Commissioner of your local ward office of your Municipality.

@vishalsodani
vishalsodani / imgproxy.md
Last active October 15, 2023 13:32
Img proxy
@vishalsodani
vishalsodani / caddy_debian.md
Created October 13, 2023 09:38
caddy with django on debian 12
  • sudo apt install caddy -y
  • sudo setcap 'cap_net_bind_service=+ep' /usr/bin/caddy
  • sudo systemctl start caddy
  • enable your service to start on boot with the command below: sudo systemctl enable caddy
@vishalsodani
vishalsodani / notes from gpt 3 paper.md
Last active September 27, 2023 06:17
notes from gpt 3 paper

Few-Shot (FS) is the term we will use in this work to refer to the setting where the model is given a few demonstrations of the task at inference time as conditioning [RWC+19 ], but no weight updates are allowed. As shown in Figure 2.1, for a typical dataset an example has a context and a desired completion (for example an English sentence and the French translation), and few-shot works by giving K examples of context and completion, and then one final example of context, with the model expected to provide the completion

GPT 3 use straining method described in this paper Language Models are Unsupervised Multitask Learners - https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf

All models were trained on V100 GPU’s on part of a high-bandwidth cluster provided by Microsoft

@vishalsodani
vishalsodani / standfor nlu part 1 notes.md
Created September 23, 2023 12:15
standfor nlu part 1 notes

how to inject truth? self consistent

  • prompting - putting a model in temporary state and generate sample from that model

    • in context learning
    • gpt 3 paper and rise of context learning
  • chain of thought prompting

@vishalsodani
vishalsodani / early_humans.md
Created September 1, 2023 08:10
early_humans

mtDNA - Mitchrondia from mother

Y chromosome from father

mutations are imp

  • mutation rate
  • haplogroup - single in Greek
  • one haplogroup L3 found in all humans mtDNA, descendant lineage of L3 are N and M ( sub haplogroups), R is subgroup of N
  • haplogroup CT for Y-chromosome
  • why date for moden humans OoA abt 70000 yrs ago; before that glacial period which was very cold to move out
@vishalsodani
vishalsodani / htmx_patterns.md
Created August 28, 2023 11:22
htmx patterns

return error and and reswap target; this is nice. no need for client side javascript

        error = render_to_string('product_info_error.html',{'error':"test"})
        response = HttpResponse(error)
        response["HX-Retarget"] = "#validation_error"
        response["HX-Reswap"] = "#innerHTML"
        return response