nebulous 0.1.86

A globally distributed container orchestrator
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49

kind: Container
metadata:
  name: ms-swift-test
  namespace: test
  labels:
    this: that
image: "us-docker.pkg.dev/agentsea-dev/orign/swift-trainer:latest"
command: |
  CUDA_VISIBLE_DEVICES=0 \
  MAX_PIXELS=1003520 \
  swift sft \
      --model Qwen/Qwen2-VL-7B-Instruct \
      --dataset 'modelscope/coco_2014_caption:train#300' \
      --train_type lora \
      --torch_dtype bfloat16 \
      --num_train_epochs 1 \
      --per_device_train_batch_size 1 \
      --per_device_eval_batch_size 1 \
      --learning_rate 1e-4 \
      --lora_rank 8 \
      --lora_alpha 32 \
      --target_modules all-linear \
      --freeze_vit true \
      --gradient_accumulation_steps 16 \
      --eval_steps 100 \
      --save_steps 100 \
      --save_total_limit 5 \
      --logging_steps 5 \
      --max_length 2048 \
      --output_dir /nebu/swift-test \
      --warmup_ratio 0.05 \
      --dataloader_num_workers 4
platform: runpod
env:
  - key: HELLO
    value: world
volumes:  
  - source: /nebu/swift-test
    dest: s3://nebulous-rs/swift-test
    driver: RCLONE_SYNC
    continuous: true
accelerators:
  - "1:A100_SXM"
meters:
  - cost: 0.1
    unit: second
    metric: runtime
    currency: USD
restart: Never