Gabriel Bianconi

`keras_preprocessing` has a new interface for `load_img`.

NTP offset definition in book and source code is incorrect

1

In https://github.com/rust-in-action/code/blob/1st-edition/ch9/ch9-clock3/src/main.rs#L40-L50: ```rust impl NTPResult { fn offset(&self) -> i64 { let delta = self.delay(); delta.abs() / 2 } fn delay(&self) -> i64 { let duration = (self.t4 - self.t1)...

Fix comments in extended CHIP-8 emulator

1

Confirmed correctness [here](http://www.cs.columbia.edu/~sedwards/classes/2016/4840-spring/designs/Chip8.pdf). - `0x7xkk`: "adds", not "sets" - S(N)E: "skip", not "store"

Conversion from logits to probabilities happens on a batch by batch basis

7

## 🐛 Bug I'm using `BinaryAUROC` and calling `.update(logits, y)` after each batch (`validation_step`), expecting the following behavior: "If preds has values outside [0,1] range we consider the input to...

[Bug]: `logprobs` is not compatible with the OpenAI spec

1

### Your current environment I'm using Runpod Serverless vLLM (https://github.com/runpod-workers/worker-vllm) so I can't run this command. However, I confirmed that the issue is in the codebase in `main`: https://github.com/vllm-project/vllm/blob/0fca3cdcf265cd375bca684d951702b6b7adf65a/vllm/entrypoints/openai/protocol.py ###...

Please verify the endpoint with the API specs. It should start with `/deposit/` (rather than `/`).

Right now the data on ClickHouse requires the config file to reconstruct the original inference and so on. Maybe we should add enough data to ClickHouse to make entries self-sufficient.

Ollama is OpenAI-compatible so it should work out of the box using the OpenAI provider (by overriding the base URL). We should explicitly test and document it to ensure there...