Today I tried to load the model weights file after torch.compile(), and then I found the weights is stored by OrderedDict data structure. This data structure is an ordered sequence, in other word its content needs to follow the order.
Neo4j is a graph database. Unlike traditional databases, the focus of a graph database is the “graph”, which means the relationships and connections between nodes (entities). Each node can represent an object (such as people, things, places…), and edges represent the relationships between nodes (such as friends, owns, located at…)
Read More »Installation Of Neo4j Graph Database And Use It By Python
Although Linux is reputed to be less susceptible to viruses distributed by hackers, it still remains a highly valuable and economically efficient operating system, particularly as a stable server or development environment. Consequently, Linux is not immune to malware or hacker attacks, albeit less frequent than in Windows systems.
Read More »[Linux] Installation And Use Of Open Source Anti-virus Tool: ClamAV Swap space plays an important role in Linux operating systems. It can be understood as a virtual memory space, where its name implies the act of swapping. Swap allows for temporarily writing less frequently used data to a fixed hard disk space to ensure stable system operation when the physical memory (RAM) is insufficient — this space is Swap.
Read More »[Linux] Increase Swap Space Using Commands to Prevent Memory Exhaustion yield is a keyword in Python, utilized much like return, the use of yield transforms a function into a Generator.
FastAPI is a modern, high-performance Python web framework. It is based on Python 3.7+ and utilizes Python type hints.
Read More »[Python] Notes on Backend Development Using the FastAPI Framework duf is a tool that offers a more “graphical representation” for checking disk space usage on Linux systems, with its name derived from Disk Usage/Free Utility. It’s worth mentioning that this tool can also be installed not only on Linux but also on MacOS and Windows.
In the midst of exploring various AI models, perhaps owing to a habit cultivated during my time in research institutes, I find myself invariably sifting through research papers pertinent to the models, systems, or libraries I’m working on.
Read More »Consensus: Harnessing AI-Powered Engines for Streamlined Research Paper Searches Recently, I trace many codes about LLaMA LoRA training (almost two months ago…), and when I reading the codes, I found if they want to configure the prompt for LLMs, if they need to change the prompt content by different training, the format_map() function is the best choice.
When using the transformers package, often we receive the following error message when we utilize different task models with various heads like AutoModelForSequenceClassification, AutoModelForSeq2SeqLM…
Read More »[Solved] Some weights of the model checkpoint at distilbert-base-multilingual-cased were not used when initializing DistilBertForSequenceClassification: [‘vocab_projector.bias’, ‘vocab_layer_norm.bias’, ‘vocab_layer_norm.weight’, ‘vocab_transform.weight’, ‘vocab_transform.bias’]
