Y por aquí el enlace al paper, por si queréis ojear más ejemplos y evaluaciones en diferentes tareas. [PAPER] https://
arxiv.org/pdf/2302.13971
.pdf
… ¿Y no molaría tener acceso libre al modelo? ¿Un GPT-3 que puedas ejecutar en local? Pues lo es Más o menos…
LLMS
-
Open-source LLM alternative to GPT-3 for local execution
By
–
-
Meta’s LLaMA Language Model Leaked After Research Release
By
–
LA FILTRACIÓN DE LLaMA… Hace unos días Meta anunció un nuevo modelo del lenguaje tipo GPT-3, que siendo de menor tamaño (65B parameters), mostraba un rendimiento similar en muchas tareas. Y Meta quiso dar acceso libre a los investigadores. Y se ha filtrado… [1/n]
-
LLaMA model performance comparable to GPT-3 without massive scale
By
–
Por dar contexto. Este modelo LLaMA es interesante porque sin ser tan masivo, resuelve muchas tareas a un rendimiento similar a GPT-3. En esencia es similar al proyecto de Chinchilla de DeepMind (70B). Aquí un ejemplo de generación de texto para que veáis que tan competente es.
-
MIT Researchers Train Logic-Aware Models Against Bias
By
–
Ask a LLM if they “think” they’re biased. Guess what they’ll say. MIT researchers trained logic-aware language models to mitigate stereotypical sentence reasoning bias using textual entailment: https://
bit.ly/3IR4OTK -
Daily Python Machine Learning and Language Models Content Sharing
By
–
That's a wrap. Every day, I share and simplify complex concepts around Python, Machine Learning & Language Models. Follow me → @Sumanth_077 if you haven't already to ensure you don't miss that. Like/RT the first tweet to support my work and help this reach more people.
-
RLHF Resources: OpenAI and Hugging Face Learning Materials
By
–
Here are some resources which you can further research on RLHF: Open AI's blog: http://
openai.com/blog/instructi
on-following/
…… Hugging Face blog: https://
huggingface.co/blog/rlhf -
SFT Model Training with Reward Function Iterations
By
–
Step3: Sample a Prompt from the dataset, give that to the SFT model and get the generated response. Use the RF model to calculate the reward of the response and use that to update the Model. Iterate over it. That's how they have developed it. 9/9
-
Using Reward Models and PPO to Fine-tune Language Models
By
–
Now we don't need Humans to rank the responses. We have RM which will take care of that. The final step is to use this RM as a reward function and fine-tune the SFT model to maximize the reward using Proximal Policy Optimization which is Reinforcement Learning Algorithm 8/9
-
Training Reward Models with Human-Ranked SFT Responses
By
–
Step2: They used the SFT model to generate multiple responses to a given prompt and the Humans ranks the responses from best to worst. Now we have a labeled dataset, then training a Reward Model will learn how the human would actually rank the responses 7/9
-
Supervised Fine-Tuning GPT-3.5 with Human-Written Answers
By
–
Step1: They have prepared a dataset of human-written answers to the prompts and used that to train the model which is called Supervised Fine-Tuning (SFT) model. The model used here for fine-tuning is the Instruct GPT(GPT-3.5) 5/9