muellerzr

# Copyright © 2024 Apple Inc.

"""
Run with:

```
OMPI_MCA_mpi_abort_print_stack=1 mlx.launch --hostfile hosts.json --backend mpi --mpi-arg "--mca btl_tcp_if_exclude lo0,169.254.0.0/16,fe80::/10 --mca mpi_show_mca_params all --mca coll_base_verbose 100 --mca btl_base_verbose 100" script.py
```

Make sure you can run MLX over MPI on two hosts. For more information see the

muellerzr

[balthasar.local:04124] btl:tcp: would block, so allowing background progress
[balthasar.local:04124] btl:tcp: now connected to 192.168.68.53, process [[49579,1],1]
[balthasar.local:04124] btl:tcp: connect() to 192.168.68.53:1024 completed (complete_connect), sending connect ACK
[melchior.local:01946] btl:tcp: now connected to 192.168.68.80, process [[49579,1],0]
[melchior.local:01946] btl: tcp: attempting to connect() to [[49579,1],0] address fdcd:ee7b:759e:1744:414:41e7:3d97:1761 on port 1024
[melchior.local:01946] btl:tcp: would block, so allowing background progress
[melchior.local:01946] btl:tcp: connect() to fdcd:ee7b:759e:1744:414:41e7:3d97:1761:1024 completed (complete_connect), sending connect ACK
[balthasar.local:04124] btl:tcp: now connected to fdcd:ee7b:759e:1744:f700:ca71:db72:f19a, process [[49579,1],1]
[melchior:01946] [0] func:0   libopen-pal.80.dylib                0x000000011a4c776c opal_backtrace_buffer + 56
[melchior:01946] [1] func:1   libmpi.40.dylib                     0x000000011a69222

muellerzr

(mlx-stuff) balthasar:mlx_stuff muellerzr$ mlx.launch --hostfile hosts.json kimi_pipeline.py 
Traceback (most recent call last):
  File "/Users/muellerzr/mlx_stuff/kimi_pipeline.py", line 82, in <module>
    model, tokenizer = shard_and_load(model_path)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/muellerzr/mlx_stuff/kimi_pipeline.py", line 56, in shard_and_load
    tokenizer = load_tokenizer(
                ^^^^^^^^^^^^^^^
  File "/Users/muellerzr/mlx_stuff/.venv/lib/python3.11/site-packages/mlx_lm/tokenizer_utils.py", line 453, in load_tokenizer
    AutoTokenizer.from_pretrained(model_path, **tokenizer_config_extra),

muellerzr

https://raw.githubusercontent.com/StevenBlack/hosts/master/hosts
https://raw.githubusercontent.com/PolishFiltersTeam/KADhosts/master/KADhosts.txt
https://raw.githubusercontent.com/FadeMind/hosts.extras/master/add.Spam/hosts
https://v.firebog.net/hosts/static/w3kbl.txt
https://adaway.org/hosts.txt
https://v.firebog.net/hosts/AdguardDNS.txt
https://v.firebog.net/hosts/Admiral.txt
https://raw.githubusercontent.com/anudeepND/blacklist/master/adservers.txt
https://v.firebog.net/hosts/Easylist.txt
 	https://pgl.yoyo.org/adservers/serverlist.php?hostformat=hosts&showintro=0&mimetype=plaintext

muellerzr

muellerzr

[project]
name = "my_project"
version = "0.1.0"
description = "A simple Python project"
authors = [
  { name = "Your Name", email = "[email protected]" }
]
readme = "README.md"
license = { text = "MIT" }
requires-python = ">=3.8"

muellerzr

import torch
from torch.utils.data import DataLoader
from transformers import AutoTokenizer, AutoModelForCausalLM, Trainer, TrainingArguments, DataCollatorForLanguageModeling
from datasets import load_from_disk
import torch.nn.functional as F
from torch.nn import CrossEntropyLoss
import numpy as np
import logging
import time
from datetime import datetime

muellerzr

import json
import os
from datetime import datetime
import argparse
from typing import Dict, List, Optional

"""
Quick Journal is a simple CLI journal that allows you to write, search, and export journal entries.
Specialized for ML experiment tracking with predefined fields.

muellerzr

import torch.nn as nn
from datasets import load_dataset
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
    DataCollatorForLanguageModeling,
    Trainer,
    TrainingArguments,
    set_seed,
)

muellerzr

# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,