keyboardAnt

PUT _watcher/watch/core_dumps_slack
{
    "trigger" : {
        "schedule" : {"interval" : "1m"}
    },
    "input" : {
        "search" : {
            "request" : {
                "body" : {
                    "query": {

keyboardAnt

keyboardAnt

keyboardAnt

keyboardAnt

# Refernce: https://static.sched.com/hosted_files/pytorch2023/c0/Accelerating%20Generative%20AI%20PTC%20%282%29.pdf?page=41

import torch


def speculative_decode(
    model: LLaMA,
    draft_model: LLaMA,
    cur_token: torch.Tensor,
    input_pos: int,

keyboardAnt

import numpy as np

# Placeholder functions to be implemented

def get_lora_adapter(username: str) -> callable:
    """Returns a callable function for the Lora adapter."""
    pass

def get_reft_adapter(username: str) -> callable:
    """Returns a callable function for the reft adapter."""

keyboardAnt

#!/usr/bin/env bash
set -euo pipefail

# Single-file repro harness for VS Code renderer crash code 5
# Usage:
#   bash single_file_repro.sh start [duration_seconds]
#   bash single_file_repro.sh status
#   bash single_file_repro.sh stop
#
# Optional:

keyboardAnt

Pairplot comparison images for PR #111

Before/after defensive groupby fix.