7/6/2025- 7/6/2025
- 7/6/2025- 7/6/2025
- 1. Datasets & DataLoaders
- 2. Utils
- 7/7/2025
- 7/8/2025
- 7/14/2025
- 7/15/2025
- 7/16/2025
- 7/17/2025
- 7/17/2025
- 7/21/2025
- 7/28/2025
I finally decided to put some efforts into learning pytorch. I am writing this notes and making this public for anyone who are interest about me.
labels_map = { // dictionary very understandable also an look up table 0: "T-Shirt", 1: "Trouser", 2: "Pullover", 3: "Dress", 4: "Coat", 5: "Sandal", 6: "Shirt", 7: "Sneaker", 8: "Bag", 9: "Ankle Boot", } figure = plt.figure(figsize=(8, 8)) // creat a canvas cols, rows = 3, 3 for i in range(1, cols * rows + 1): sample_idx = torch.randint(len(training_data), size=(1,)).item() //torch.randint(N, size=(1,)) → random integer in [0, N − 1]. //.item() converts the 1-element tensor to a plain Python int. img, label = training_data[sample_idx] figure.add_subplot(rows, cols, i) plt.title(labels_map[label]) plt.axis("off") plt.imshow(img.squeeze(), cmap="gray") plt.show() train_dataset, val_dataset = random_split(dataset, [80, 20]) train_loader = DataLoader(dataset=train_dataset, batch_size=16) val_loader = DataLoader(dataset=val_dataset, batch_size=20) *loading one batch for each loop
too many funcitons, not so interesting. Found another interesting blog, https://huggingface.co/blog/dvgodoy/beginner-pytorch-tutorial?utm_source=chatgpt.com.
learned about the difference between GPU tensor and CPU tensor(Numpy) " In PyTorch, every method that ends with an underscore (_) makes changes in-place, meaning, they will modify the underlying variable. " torch.no_grad() : tell PyTorch to “back off” and let us update our parameters without messing up with its fancy dynamic computation graph
finished my first ML competiton on Kaggle
some C++ learning notes: std:vector do not have push_front, I can only use insert(v.begin(), obj)
some advanced usage of sort: assending order based on the second element
sort(points.begin(), points.end(), [](const vector& a, const vector& b) { return a[1] < b[1]; });
- I resubmited two codes for the titanic competition and scored around 0.8, a huge improvement from what I did from the last time
- BERT for beginner
- Token embeddings: A [CLS] token is added to the input word tokens at the beginning of the first sentence and a [SEP] token is inserted at the end of each sentence.
- Segment embeddings:A marker indicating Sentence A or Sentence B is added to each token. This allows the encoder to distinguish between sentences.
- Positional embeddings:A positional embedding is added to each token to indicate its position in the sentence.
I was trying to use a List in CPP as a stack, but I found out that I could only use std:vector or std:stack: #include
std::vector stack; add elements：stack.push_back('('); pop elements：stack.pop_back(); visit top：stack.back();
#include
std::stack stack;
add elements：stack.push('('); pop elements：stack.pop(); visit top：stack.top();
Encountered one problem in Leetcode. The questions was initially asking me to simplify a path, but I found spliting a string is kind of a problem in C++. As far as I know, it is easy to do it in Python using split funciton.
I learned that I can use stringstream ss(path) and getline function to achieve the same thing: std::getline(istream& is, string& str); std::getline(istream& is, string& str, char delim);
start self-learning CS336
Last week, I finished implementing a BPE training function, and a tokenizer class still awaits me to build. Fun but a bit tough task for me.
I would like to make some notes of the difference between C++ and Python as I made those errors in coding questions C++ do not use : to star a block, this mistake sounds insane but I actually typed it when coding C++ use try and catch block, but do not use try and except It is not valid to use vec[-1] for accessing the last element in an vector in C++