aitextgen
AI
text generation library
SHOW HN2020
How HN took it
8 positive, 13 neutral or asking, 0 mixed, 3 critical, out of 24 comments about the project. Thread score +0.24.
Docs: 1 for, 0 againstExecution: 0 for, 1 againstDesign: 1 for, 0 againstPerformance: 0 for, 1 againstIdea: 0 for, 1 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Clean, developer-friendly API and usability
From 2 comments
Your API looks really clean but what's the difference between this and just GPT-2 (or) HuggingFace's implementations?
alphagrep12345 on HNI've been following minimaxir's work with GPT-2 for a while - I've tried building things on https://github.com/minimaxir/gpt-2-simple for example - and this looks like a HUGE…
simonw on HN
Well-written, newbie-friendly guide
From 1 comment
This is just brilliant. For someone who has little working knowledge but has massive interest in this field I found your guide exceptionally well written and newbie friendly (the…
superasn on HN
Faster generation and better memory efficiency than gpt-2-simple
From 1 comment
Huge fan of your gpt2-simple library, which I used to train a satirical news generator in a Colab notebook:…
neoncontrails on HN
Objections
Unclear how it differs from HuggingFace Transformers
From 2 comments
Your API looks really clean but what's the difference between this and just GPT-2 (or) HuggingFace's implementations?
alphagrep12345 on HNOne question: Why put more sh!!t on top of huggingface already really good code?
master_yoda_1 on HN
Fine-tuning large models and large datasets not yet supported
From 2 comments
> Do these memory management improvements make it possible to fine-tune a larger one? Unfortunately not yet; I need to implement gradient checkpointing first. Memory-wise, the…
minimaxir on HN~50-100G isn't "some" text data. The original GPT-2 was trained on 40G of text data. I'm not 100% sure you can encode and store that much data in memory with the current…
minimaxir on HN
Requests
Gradient checkpointing for fine-tuning larger models
From 2 comments
Huge fan of your gpt2-simple library, which I used to train a satirical news generator in a Colab notebook:…
neoncontrails on HNHave I been using gpt-2-simple wrong..? I’ve been fine-tuning 355M on a 8GB 1080 for months..
britmob on HN
Support for parentheses in the prompt argument
From 1 comment
I get a source.error("unbalanced parenthesis") if I put unbalanced parentheses in the --prompt "argument)"
totetsu on HN
Questions
How to download the models
From 1 comment
how does one download these models?
totetsu on HN
Whether training data size of 50-100G is feasible
From 1 comment
This looks great! If I want to fine-tune this to some text data, are there obvious constraints to be aware of? I've got a reasonable amount of text (~50-100G) but seeing that…
IanCal on HN
Whether the model can be embedded on a Raspberry Pi
From 1 comment
Does anyone know an efficient way to "embed" models like this? I'm currently working in a Tamagotchi-style RPI toy and I use GPT-2 to generate answers to the chat. I wrote a…
jakearmitage on HN
Alternatives named
gpt-2-simple
From 2 comments
Huge fan of your gpt2-simple library, which I used to train a satirical news generator in a Colab notebook:…
neoncontrails on HNI've been following minimaxir's work with GPT-2 for a while - I've tried building things on https://github.com/minimaxir/gpt-2-simple for example - and this looks like a HUGE…
simonw on HN
HuggingFace Transformers
From 1 comment
Your API looks really clean but what's the difference between this and just GPT-2 (or) HuggingFace's implementations?
alphagrep12345 on HN