trainwave-cli 0.4.2__tar.gz → 0.4.3__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/.gitlab-ci.yml +8 -1
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/PKG-INFO +15 -14
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/basic/trainwave.toml +6 -4
- trainwave_cli-0.4.3/example/pytorch-llm/requirements.txt +8 -0
- trainwave_cli-0.4.3/example/pytorch-llm/run.sh +5 -0
- trainwave_cli-0.4.3/example/pytorch-llm/setup.sh +5 -0
- trainwave_cli-0.4.3/example/pytorch-llm/train.py +164 -0
- trainwave_cli-0.4.3/example/pytorch-llm/trainwave.toml +10 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/pyproject.toml +24 -21
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/ruff.toml +2 -0
- trainwave_cli-0.4.3/tests/cli/test_file_secrets.py +102 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/cli/test_jobs.py +156 -1
- trainwave_cli-0.4.3/tests/conftest.py +10 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/test_api.py +39 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/api.py +82 -10
- trainwave_cli-0.4.3/trainwave_cli/cli/file_secrets.py +123 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/jobs.py +33 -3
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/logs.py +4 -3
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/ssh_keys.py +3 -1
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/io.py +9 -5
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/main.py +13 -5
- trainwave_cli-0.4.3/uv.lock +2419 -0
- trainwave_cli-0.4.2/tests/conftest.py +0 -21
- trainwave_cli-0.4.2/uv.lock +0 -1772
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/.dockerignore +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/.gitignore +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/Makefile +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/README.md +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/README.md +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/basic/requirements.txt +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/basic/run.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/basic/setup.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/basic/train.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/disk-exhaustion/requirements.txt +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/disk-exhaustion/run.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/disk-exhaustion/setup.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/disk-exhaustion/train.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/disk-exhaustion/trainwave.toml +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/memory-exhaustion/requirements.txt +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/memory-exhaustion/run.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/memory-exhaustion/setup.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/memory-exhaustion/train.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/memory-exhaustion/trainwave.toml +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/secret-leak/requirements.txt +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/secret-leak/run.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/secret-leak/setup.sh +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/secret-leak/train.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/example/secret-leak/trainwave.toml +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/pyrightconfig.json +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/cli/test_auth.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/cli/test_secrets.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/cli/test_setup.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/test_io.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/test_main.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/tests/test_utils.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/__init__.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/auth.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/offers.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/secrets.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/cli/setup.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/config/config.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/format_utils.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/utils.py +0 -0
- {trainwave_cli-0.4.2 → trainwave_cli-0.4.3}/trainwave_cli/xceptions.py +0 -0
|
@@ -20,6 +20,13 @@ publish:
|
|
|
20
20
|
- curl -LsSf https://astral.sh/uv/install.sh | sh
|
|
21
21
|
- export PATH="/root/.local/bin:$PATH"
|
|
22
22
|
- uv build
|
|
23
|
-
|
|
23
|
+
# Only publish when the version in pyproject.toml has been bumped
|
|
24
|
+
- |
|
|
25
|
+
VERSION=$(sed -n 's/^version = "\(.*\)"/\1/p' pyproject.toml)
|
|
26
|
+
if curl -fsS "https://pypi.org/pypi/trainwave-cli/$VERSION/json" > /dev/null; then
|
|
27
|
+
echo "trainwave-cli $VERSION is already on PyPI, skipping publish"
|
|
28
|
+
else
|
|
29
|
+
uv publish --token $PYPI_API_KEY
|
|
30
|
+
fi
|
|
24
31
|
only:
|
|
25
32
|
- main
|
|
@@ -1,6 +1,6 @@
|
|
|
1
|
-
Metadata-Version: 2.
|
|
1
|
+
Metadata-Version: 2.5
|
|
2
2
|
Name: trainwave-cli
|
|
3
|
-
Version: 0.4.
|
|
3
|
+
Version: 0.4.3
|
|
4
4
|
Summary: Trainwave CLI - A command-line interface for managing Trainwave resources
|
|
5
5
|
Project-URL: Homepage, https://trainwave.ai
|
|
6
6
|
Project-URL: Documentation, https://trainwave.ai/docs
|
|
@@ -21,22 +21,23 @@ Classifier: Programming Language :: Python :: 3.12
|
|
|
21
21
|
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
22
22
|
Classifier: Topic :: Software Development :: Libraries :: Python Modules
|
|
23
23
|
Requires-Python: >=3.10
|
|
24
|
-
Requires-Dist: aiofiles<25
|
|
25
|
-
Requires-Dist: aiohttp<4,>=3.
|
|
26
|
-
Requires-Dist: beaupy<4,>=3.
|
|
27
|
-
Requires-Dist: boto3>=1.
|
|
24
|
+
Requires-Dist: aiofiles<26,>=25.1.0
|
|
25
|
+
Requires-Dist: aiohttp<4,>=3.14
|
|
26
|
+
Requires-Dist: beaupy<4,>=3.12.0
|
|
27
|
+
Requires-Dist: boto3>=1.43
|
|
28
28
|
Requires-Dist: dacite<2,>=1.9.2
|
|
29
29
|
Requires-Dist: httpx==0.28.1
|
|
30
|
-
Requires-Dist: human-readable<2
|
|
31
|
-
Requires-Dist: loguru<1,>=0.7.
|
|
32
|
-
Requires-Dist: pathspec<1
|
|
30
|
+
Requires-Dist: human-readable<3,>=2.0.3
|
|
31
|
+
Requires-Dist: loguru<1,>=0.7.3
|
|
32
|
+
Requires-Dist: pathspec<2,>=1.1.1
|
|
33
33
|
Requires-Dist: pytimeparse<2,>=1.1.8
|
|
34
|
-
Requires-Dist: tabulate<1,>=0.
|
|
34
|
+
Requires-Dist: tabulate<1,>=0.10.0
|
|
35
35
|
Requires-Dist: toml<1,>=0.10.2
|
|
36
|
-
Requires-Dist: tqdm<5,>=4.
|
|
37
|
-
Requires-Dist: typer-config[all]<2,>=1.
|
|
38
|
-
Requires-Dist: typer<1,>=0.
|
|
39
|
-
Requires-Dist: websockets<
|
|
36
|
+
Requires-Dist: tqdm<5,>=4.69.0
|
|
37
|
+
Requires-Dist: typer-config[all]<2,>=1.5.1
|
|
38
|
+
Requires-Dist: typer<1,>=0.27.0
|
|
39
|
+
Requires-Dist: websockets<17,>=16.1.1; python_version < '3.11'
|
|
40
|
+
Requires-Dist: websockets<18,>=17.1; python_version >= '3.11'
|
|
40
41
|
Description-Content-Type: text/markdown
|
|
41
42
|
|
|
42
43
|
# Trainwave CLI
|
|
@@ -2,13 +2,15 @@ name = "Example - Basic training"
|
|
|
2
2
|
expires = "1h"
|
|
3
3
|
setup_command = "bash setup.sh"
|
|
4
4
|
run_command = "bash run.sh"
|
|
5
|
-
image = "trainwave/pytorch:2.4.0"
|
|
5
|
+
# image = "trainwave/pytorch:2.4.0"
|
|
6
6
|
framework = "PyTorch"
|
|
7
7
|
project = "p-kgzxmqte"
|
|
8
|
-
compliance_soc2 = true
|
|
9
|
-
gpus = 1
|
|
10
|
-
gpu_type = "NVIDIA-A40-48GB"
|
|
8
|
+
# compliance_soc2 = true
|
|
9
|
+
# gpus = 1
|
|
10
|
+
# gpu_type = "NVIDIA-A40-48GB"
|
|
11
11
|
organization = "o-kgzxmqte"
|
|
12
|
+
cpus = 1
|
|
13
|
+
burst_count = 3
|
|
12
14
|
|
|
13
15
|
[env_vars]
|
|
14
16
|
TESTING1 = "1"
|
|
@@ -0,0 +1,164 @@
|
|
|
1
|
+
"""
|
|
2
|
+
Minimal but realistic LLM-training-shaped demo:
|
|
3
|
+
|
|
4
|
+
- Random-init Hugging Face GPT-2 (tiny config, no weight download)
|
|
5
|
+
- Parameter-efficient fine-tuning via PEFT / LoRA
|
|
6
|
+
- `datasets` for batched sequences
|
|
7
|
+
- `accelerate` for device placement and mixed precision prep
|
|
8
|
+
"""
|
|
9
|
+
|
|
10
|
+
from __future__ import annotations
|
|
11
|
+
|
|
12
|
+
import json
|
|
13
|
+
import math
|
|
14
|
+
import os
|
|
15
|
+
import random
|
|
16
|
+
from pathlib import Path
|
|
17
|
+
|
|
18
|
+
import torch
|
|
19
|
+
from accelerate import Accelerator
|
|
20
|
+
from accelerate.utils import set_seed
|
|
21
|
+
from datasets import Dataset
|
|
22
|
+
from peft import LoraConfig, TaskType, get_peft_model
|
|
23
|
+
from torch.optim import AdamW
|
|
24
|
+
from torch.utils.data import DataLoader
|
|
25
|
+
from tqdm.auto import tqdm
|
|
26
|
+
from transformers import GPT2Config, GPT2LMHeadModel, get_cosine_schedule_with_warmup
|
|
27
|
+
|
|
28
|
+
|
|
29
|
+
def _synthetic_causal_batch(
|
|
30
|
+
*,
|
|
31
|
+
num_rows: int,
|
|
32
|
+
seq_len: int,
|
|
33
|
+
vocab_size: int,
|
|
34
|
+
seed: int,
|
|
35
|
+
) -> Dataset:
|
|
36
|
+
"""Token ids in [0, vocab_size); looks like LM data without touching the network."""
|
|
37
|
+
rng = random.Random(seed)
|
|
38
|
+
rows: list[list[int]] = []
|
|
39
|
+
for _ in range(num_rows):
|
|
40
|
+
rows.append([rng.randrange(vocab_size) for _ in range(seq_len)])
|
|
41
|
+
return Dataset.from_dict({"input_ids": rows})
|
|
42
|
+
|
|
43
|
+
|
|
44
|
+
def collate_tokens(
|
|
45
|
+
features: list[dict[str, list[int]]],
|
|
46
|
+
) -> dict[str, torch.Tensor]:
|
|
47
|
+
batch = torch.tensor([f["input_ids"] for f in features], dtype=torch.long)
|
|
48
|
+
return {"input_ids": batch}
|
|
49
|
+
|
|
50
|
+
|
|
51
|
+
def main() -> None:
|
|
52
|
+
out_dir = Path(os.environ.get("TRAINWAVE_OUTPUT_DIR", "/workspace/output"))
|
|
53
|
+
out_dir.mkdir(parents=True, exist_ok=True)
|
|
54
|
+
|
|
55
|
+
# Small GPT-2-shaped LM: fast on CPU, still exercises the HF stack.
|
|
56
|
+
vocab_size = 512
|
|
57
|
+
config = GPT2Config(
|
|
58
|
+
vocab_size=vocab_size,
|
|
59
|
+
n_positions=128,
|
|
60
|
+
n_embd=128,
|
|
61
|
+
n_layer=4,
|
|
62
|
+
n_head=4,
|
|
63
|
+
n_inner=512,
|
|
64
|
+
resid_pdrop=0.0,
|
|
65
|
+
embd_pdrop=0.0,
|
|
66
|
+
attn_pdrop=0.0,
|
|
67
|
+
)
|
|
68
|
+
model = GPT2LMHeadModel(config)
|
|
69
|
+
|
|
70
|
+
peft_config = LoraConfig(
|
|
71
|
+
task_type=TaskType.CAUSAL_LM,
|
|
72
|
+
r=8,
|
|
73
|
+
lora_alpha=16,
|
|
74
|
+
lora_dropout=0.05,
|
|
75
|
+
target_modules=["c_attn", "c_proj"],
|
|
76
|
+
fan_in_fan_out=True, # GPT-2 attention uses Conv1D
|
|
77
|
+
)
|
|
78
|
+
model = get_peft_model(model, peft_config)
|
|
79
|
+
|
|
80
|
+
mp = "fp16" if torch.cuda.is_available() else "no"
|
|
81
|
+
accelerator = Accelerator(mixed_precision=mp)
|
|
82
|
+
set_seed(42)
|
|
83
|
+
device = accelerator.device
|
|
84
|
+
|
|
85
|
+
dataset = _synthetic_causal_batch(
|
|
86
|
+
num_rows=256,
|
|
87
|
+
seq_len=config.n_positions,
|
|
88
|
+
vocab_size=vocab_size,
|
|
89
|
+
seed=42,
|
|
90
|
+
)
|
|
91
|
+
loader = DataLoader(
|
|
92
|
+
dataset,
|
|
93
|
+
batch_size=8,
|
|
94
|
+
shuffle=True,
|
|
95
|
+
collate_fn=collate_tokens,
|
|
96
|
+
num_workers=0,
|
|
97
|
+
)
|
|
98
|
+
|
|
99
|
+
train_steps = 50
|
|
100
|
+
optimizer = AdamW(model.parameters(), lr=3e-4, weight_decay=0.01)
|
|
101
|
+
scheduler = get_cosine_schedule_with_warmup(
|
|
102
|
+
optimizer,
|
|
103
|
+
num_warmup_steps=5,
|
|
104
|
+
num_training_steps=train_steps,
|
|
105
|
+
)
|
|
106
|
+
|
|
107
|
+
model, optimizer, loader, scheduler = accelerator.prepare(
|
|
108
|
+
model, optimizer, loader, scheduler
|
|
109
|
+
)
|
|
110
|
+
|
|
111
|
+
model.train()
|
|
112
|
+
losses: list[float] = []
|
|
113
|
+
pbar = tqdm(range(train_steps), desc="train", disable=not accelerator.is_local_main_process)
|
|
114
|
+
data_iter = iter(loader)
|
|
115
|
+
|
|
116
|
+
for step in pbar:
|
|
117
|
+
try:
|
|
118
|
+
batch = next(data_iter)
|
|
119
|
+
except StopIteration:
|
|
120
|
+
data_iter = iter(loader)
|
|
121
|
+
batch = next(data_iter)
|
|
122
|
+
|
|
123
|
+
batch = {k: v.to(device) for k, v in batch.items()}
|
|
124
|
+
outputs = model(**batch, labels=batch["input_ids"])
|
|
125
|
+
loss = outputs.loss
|
|
126
|
+
|
|
127
|
+
accelerator.backward(loss)
|
|
128
|
+
optimizer.step()
|
|
129
|
+
scheduler.step()
|
|
130
|
+
optimizer.zero_grad()
|
|
131
|
+
|
|
132
|
+
loss_val = float(accelerator.reduce(loss.detach(), reduction="mean"))
|
|
133
|
+
losses.append(loss_val)
|
|
134
|
+
pbar.set_postfix(loss=f"{loss_val:.4f}")
|
|
135
|
+
|
|
136
|
+
summary = {
|
|
137
|
+
"device": str(device),
|
|
138
|
+
"dtype_policy": str(accelerator.mixed_precision),
|
|
139
|
+
"trainable_params": sum(p.numel() for p in model.parameters() if p.requires_grad),
|
|
140
|
+
"total_params": sum(p.numel() for p in model.parameters()),
|
|
141
|
+
"final_loss": losses[-1] if losses else None,
|
|
142
|
+
"mean_loss_last_10": sum(losses[-10:]) / min(10, len(losses)) if losses else None,
|
|
143
|
+
}
|
|
144
|
+
|
|
145
|
+
if accelerator.is_main_process:
|
|
146
|
+
# `safetensors` is expected by many HF workflows; save adapter weights only.
|
|
147
|
+
adapter_dir = out_dir / "lora_adapter"
|
|
148
|
+
model.save_pretrained(adapter_dir)
|
|
149
|
+
|
|
150
|
+
with (out_dir / "metrics.json").open("w", encoding="utf-8") as f:
|
|
151
|
+
json.dump(summary, f, indent=2)
|
|
152
|
+
|
|
153
|
+
print(
|
|
154
|
+
"Done. "
|
|
155
|
+
f"Trainable / total params: {summary['trainable_params']:,} / "
|
|
156
|
+
f"{summary['total_params']:,}. "
|
|
157
|
+
f"Final loss ≈ {summary['final_loss']:.4f} "
|
|
158
|
+
f"(random data; expect high but finite CE ≈ ln({vocab_size}) ≈ "
|
|
159
|
+
f"{math.log(vocab_size):.2f})."
|
|
160
|
+
)
|
|
161
|
+
|
|
162
|
+
|
|
163
|
+
if __name__ == "__main__":
|
|
164
|
+
main()
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
[project]
|
|
2
2
|
name = "trainwave-cli"
|
|
3
|
-
version = "0.4.
|
|
3
|
+
version = "0.4.3"
|
|
4
4
|
description = "Trainwave CLI - A command-line interface for managing Trainwave resources"
|
|
5
5
|
authors = [{ name = "Johan Backman", email = "johan@trainwave.ai" }]
|
|
6
6
|
maintainers = [{ name = "Johan Backman", email = "johan@trainwave.ai" }]
|
|
@@ -23,22 +23,24 @@ classifiers = [
|
|
|
23
23
|
"Topic :: Software Development :: Libraries :: Python Modules",
|
|
24
24
|
]
|
|
25
25
|
dependencies = [
|
|
26
|
-
"aiofiles>=
|
|
27
|
-
"aiohttp>=3.
|
|
28
|
-
"beaupy>=3.
|
|
29
|
-
"boto3>=1.
|
|
26
|
+
"aiofiles>=25.1.0,<26",
|
|
27
|
+
"aiohttp>=3.14,<4",
|
|
28
|
+
"beaupy>=3.12.0,<4",
|
|
29
|
+
"boto3>=1.43",
|
|
30
30
|
"dacite>=1.9.2,<2",
|
|
31
31
|
"httpx==0.28.1",
|
|
32
|
-
"human-readable>=
|
|
33
|
-
"loguru>=0.7.
|
|
34
|
-
"pathspec>=
|
|
32
|
+
"human-readable>=2.0.3,<3",
|
|
33
|
+
"loguru>=0.7.3,<1",
|
|
34
|
+
"pathspec>=1.1.1,<2",
|
|
35
35
|
"pytimeparse>=1.1.8,<2",
|
|
36
|
-
"tabulate>=0.
|
|
36
|
+
"tabulate>=0.10.0,<1",
|
|
37
37
|
"toml>=0.10.2,<1",
|
|
38
|
-
"tqdm>=4.
|
|
39
|
-
"typer>=0.
|
|
40
|
-
"typer-config[all]>=1.
|
|
41
|
-
|
|
38
|
+
"tqdm>=4.69.0,<5",
|
|
39
|
+
"typer>=0.27.0,<1",
|
|
40
|
+
"typer-config[all]>=1.5.1,<2",
|
|
41
|
+
# websockets 17 dropped Python 3.10; keep 3.10 on the last compatible release.
|
|
42
|
+
"websockets>=17.1,<18; python_version >= '3.11'",
|
|
43
|
+
"websockets>=16.1.1,<17; python_version < '3.11'",
|
|
42
44
|
]
|
|
43
45
|
|
|
44
46
|
[project.urls]
|
|
@@ -50,15 +52,16 @@ wave = "trainwave_cli.main:entrypoint"
|
|
|
50
52
|
|
|
51
53
|
[dependency-groups]
|
|
52
54
|
dev = [
|
|
53
|
-
"freezegun>=1.5.
|
|
55
|
+
"freezegun>=1.5.5,<2",
|
|
54
56
|
"ipdb>=0.13.13,<1",
|
|
55
|
-
|
|
56
|
-
"
|
|
57
|
-
"pytest
|
|
58
|
-
"pytest-
|
|
59
|
-
"pytest-
|
|
60
|
-
"pytest-
|
|
61
|
-
"
|
|
57
|
+
# ipython 9.x requires Python >=3.11; uncapped so 3.10 resolves to 8.x.
|
|
58
|
+
"ipython>=8.34.0",
|
|
59
|
+
"pytest>=9.1.1,<10",
|
|
60
|
+
"pytest-asyncio>=1.4.0,<2",
|
|
61
|
+
"pytest-cov>=7.1.0,<8",
|
|
62
|
+
"pytest-httpx>=0.36.2,<1",
|
|
63
|
+
"pytest-mock>=3.15.1,<4",
|
|
64
|
+
"ruff>=0.16.10",
|
|
62
65
|
]
|
|
63
66
|
|
|
64
67
|
[build-system]
|
|
@@ -66,6 +66,8 @@ lint.ignore = [
|
|
|
66
66
|
# Let black handle line length checking
|
|
67
67
|
"E501",
|
|
68
68
|
"PERF203",
|
|
69
|
+
# Became stable in ruff 0.16; Typer commands/API helpers legitimately take many parameters
|
|
70
|
+
"PLR0917",
|
|
69
71
|
]
|
|
70
72
|
# Allow fix for all enabled rules (when `--fix`) is provided.
|
|
71
73
|
lint.fixable = ["ALL"]
|
|
@@ -0,0 +1,102 @@
|
|
|
1
|
+
import asyncio
|
|
2
|
+
from datetime import datetime, timezone
|
|
3
|
+
from pathlib import Path
|
|
4
|
+
from unittest.mock import AsyncMock, patch
|
|
5
|
+
|
|
6
|
+
import pytest
|
|
7
|
+
from typer.testing import CliRunner
|
|
8
|
+
|
|
9
|
+
from trainwave_cli.api import InstanceFileSecret
|
|
10
|
+
from trainwave_cli.cli.file_secrets import app
|
|
11
|
+
from trainwave_cli.config.config import config
|
|
12
|
+
|
|
13
|
+
runner = CliRunner()
|
|
14
|
+
|
|
15
|
+
|
|
16
|
+
@pytest.fixture(autouse=True)
|
|
17
|
+
def setup_config():
|
|
18
|
+
config.api_key = "test-key"
|
|
19
|
+
config.endpoint = "https://api.test.com"
|
|
20
|
+
|
|
21
|
+
|
|
22
|
+
@pytest.fixture
|
|
23
|
+
def file_secret() -> InstanceFileSecret:
|
|
24
|
+
return InstanceFileSecret(
|
|
25
|
+
id="secret-id",
|
|
26
|
+
rid="fs-secret",
|
|
27
|
+
instance="instance-id",
|
|
28
|
+
name="nats credentials",
|
|
29
|
+
relative_path="dynamo/nats.creds",
|
|
30
|
+
digest="abcdef1234567890",
|
|
31
|
+
mode=0o600,
|
|
32
|
+
created_at=datetime.now(timezone.utc),
|
|
33
|
+
)
|
|
34
|
+
|
|
35
|
+
|
|
36
|
+
@pytest.mark.asyncio
|
|
37
|
+
async def test_list_file_secrets(file_secret: InstanceFileSecret) -> None:
|
|
38
|
+
api = AsyncMock()
|
|
39
|
+
api.list_instance_file_secrets.return_value = [file_secret]
|
|
40
|
+
|
|
41
|
+
with patch("trainwave_cli.cli.file_secrets.Api", return_value=api):
|
|
42
|
+
result = await asyncio.to_thread(
|
|
43
|
+
runner.invoke, app, ["list", "--job", "job-id"]
|
|
44
|
+
)
|
|
45
|
+
|
|
46
|
+
assert result.exit_code == 0
|
|
47
|
+
assert "nats credentials" in result.stdout
|
|
48
|
+
assert "dynamo/nats.creds" in result.stdout
|
|
49
|
+
assert "abcdef1234567890" in result.stdout
|
|
50
|
+
api.list_instance_file_secrets.assert_called_once_with("job-id")
|
|
51
|
+
|
|
52
|
+
|
|
53
|
+
@pytest.mark.asyncio
|
|
54
|
+
async def test_set_file_secret_reads_file_without_printing_value(
|
|
55
|
+
tmp_path: Path, file_secret: InstanceFileSecret
|
|
56
|
+
) -> None:
|
|
57
|
+
source = tmp_path / "nats.creds"
|
|
58
|
+
source.write_text("private credential material")
|
|
59
|
+
api = AsyncMock()
|
|
60
|
+
api.set_instance_file_secret.return_value = file_secret
|
|
61
|
+
|
|
62
|
+
with patch("trainwave_cli.cli.file_secrets.Api", return_value=api):
|
|
63
|
+
result = await asyncio.to_thread(
|
|
64
|
+
runner.invoke,
|
|
65
|
+
app,
|
|
66
|
+
[
|
|
67
|
+
"set",
|
|
68
|
+
"--job",
|
|
69
|
+
"job-id",
|
|
70
|
+
"--instance",
|
|
71
|
+
"instance-id",
|
|
72
|
+
"--name",
|
|
73
|
+
"nats credentials",
|
|
74
|
+
"--relative-path",
|
|
75
|
+
"dynamo/nats.creds",
|
|
76
|
+
"--file",
|
|
77
|
+
str(source),
|
|
78
|
+
],
|
|
79
|
+
)
|
|
80
|
+
|
|
81
|
+
assert result.exit_code == 0
|
|
82
|
+
assert "private credential material" not in result.stdout
|
|
83
|
+
api.set_instance_file_secret.assert_called_once_with(
|
|
84
|
+
"job-id",
|
|
85
|
+
"instance-id",
|
|
86
|
+
"nats credentials",
|
|
87
|
+
"dynamo/nats.creds",
|
|
88
|
+
"private credential material",
|
|
89
|
+
None,
|
|
90
|
+
)
|
|
91
|
+
|
|
92
|
+
|
|
93
|
+
@pytest.mark.asyncio
|
|
94
|
+
async def test_revoke_file_secret() -> None:
|
|
95
|
+
api = AsyncMock()
|
|
96
|
+
with patch("trainwave_cli.cli.file_secrets.Api", return_value=api):
|
|
97
|
+
result = await asyncio.to_thread(
|
|
98
|
+
runner.invoke, app, ["revoke", "fs-secret", "--job", "job-id"]
|
|
99
|
+
)
|
|
100
|
+
|
|
101
|
+
assert result.exit_code == 0
|
|
102
|
+
api.revoke_instance_file_secret.assert_called_once_with("job-id", "fs-secret")
|
|
@@ -3,10 +3,11 @@ from datetime import datetime, timezone
|
|
|
3
3
|
from unittest.mock import AsyncMock, patch
|
|
4
4
|
|
|
5
5
|
import pytest
|
|
6
|
+
import typer
|
|
6
7
|
from typer.testing import CliRunner
|
|
7
8
|
|
|
8
9
|
from trainwave_cli.api import CloudOffer, Job, JobConfig, JobStatus, TrainwaveProject
|
|
9
|
-
from trainwave_cli.cli.jobs import app
|
|
10
|
+
from trainwave_cli.cli.jobs import _parse_env_vars, app
|
|
10
11
|
from trainwave_cli.config.config import config
|
|
11
12
|
|
|
12
13
|
runner = CliRunner()
|
|
@@ -194,3 +195,157 @@ async def test_launch_job_with_gpu(mock_job, tmp_path):
|
|
|
194
195
|
call_args = mock_api.create_job.call_args[0][0]
|
|
195
196
|
assert call_args["gpu_type"] == "A100"
|
|
196
197
|
assert call_args["gpus"] == _expected_gpu_count
|
|
198
|
+
|
|
199
|
+
|
|
200
|
+
@pytest.mark.asyncio
|
|
201
|
+
async def test_launch_job_with_burst_count(mock_job, tmp_path):
|
|
202
|
+
_expected_burst_count = 3
|
|
203
|
+
config.api_key = "test-key"
|
|
204
|
+
|
|
205
|
+
mock_api = AsyncMock()
|
|
206
|
+
mock_api.create_job.return_value = mock_job
|
|
207
|
+
|
|
208
|
+
workspace = tmp_path / "workspace"
|
|
209
|
+
workspace.mkdir()
|
|
210
|
+
(workspace / "test.py").write_text("print('hello')")
|
|
211
|
+
|
|
212
|
+
with (
|
|
213
|
+
patch("trainwave_cli.cli.jobs.Api", return_value=mock_api),
|
|
214
|
+
patch("trainwave_cli.cli.jobs.Path.cwd", return_value=workspace),
|
|
215
|
+
):
|
|
216
|
+
result = await asyncio.to_thread(
|
|
217
|
+
runner.invoke,
|
|
218
|
+
app,
|
|
219
|
+
[
|
|
220
|
+
"launch",
|
|
221
|
+
"--name",
|
|
222
|
+
"test-job",
|
|
223
|
+
"--project",
|
|
224
|
+
"test-project",
|
|
225
|
+
"--setup-command",
|
|
226
|
+
"pip install -r requirements.txt",
|
|
227
|
+
"--run-command",
|
|
228
|
+
"python test.py",
|
|
229
|
+
"--burst-count",
|
|
230
|
+
str(_expected_burst_count),
|
|
231
|
+
],
|
|
232
|
+
)
|
|
233
|
+
|
|
234
|
+
assert result.exit_code == 0
|
|
235
|
+
call_args = mock_api.create_job.call_args[0][0]
|
|
236
|
+
assert call_args["burst_count"] == _expected_burst_count
|
|
237
|
+
|
|
238
|
+
|
|
239
|
+
@pytest.mark.asyncio
|
|
240
|
+
async def test_launch_job_omits_burst_count_by_default(mock_job, tmp_path):
|
|
241
|
+
config.api_key = "test-key"
|
|
242
|
+
|
|
243
|
+
mock_api = AsyncMock()
|
|
244
|
+
mock_api.create_job.return_value = mock_job
|
|
245
|
+
|
|
246
|
+
workspace = tmp_path / "workspace"
|
|
247
|
+
workspace.mkdir()
|
|
248
|
+
(workspace / "test.py").write_text("print('hello')")
|
|
249
|
+
|
|
250
|
+
with (
|
|
251
|
+
patch("trainwave_cli.cli.jobs.Api", return_value=mock_api),
|
|
252
|
+
patch("trainwave_cli.cli.jobs.Path.cwd", return_value=workspace),
|
|
253
|
+
):
|
|
254
|
+
result = await asyncio.to_thread(
|
|
255
|
+
runner.invoke,
|
|
256
|
+
app,
|
|
257
|
+
[
|
|
258
|
+
"launch",
|
|
259
|
+
"--name",
|
|
260
|
+
"test-job",
|
|
261
|
+
"--project",
|
|
262
|
+
"test-project",
|
|
263
|
+
"--setup-command",
|
|
264
|
+
"pip install -r requirements.txt",
|
|
265
|
+
"--run-command",
|
|
266
|
+
"python test.py",
|
|
267
|
+
],
|
|
268
|
+
)
|
|
269
|
+
|
|
270
|
+
assert result.exit_code == 0
|
|
271
|
+
call_args = mock_api.create_job.call_args[0][0]
|
|
272
|
+
assert "burst_count" not in call_args
|
|
273
|
+
|
|
274
|
+
|
|
275
|
+
async def _launch_from_toml(mock_job, tmp_path, toml_body: str):
|
|
276
|
+
"""
|
|
277
|
+
Run `wave jobs launch --config <file>` the way the integration tests do.
|
|
278
|
+
|
|
279
|
+
Every option comes from the TOML file.
|
|
280
|
+
"""
|
|
281
|
+
config.api_key = "test-key"
|
|
282
|
+
mock_api = AsyncMock()
|
|
283
|
+
mock_api.create_job.return_value = mock_job
|
|
284
|
+
|
|
285
|
+
workspace = tmp_path / "workspace"
|
|
286
|
+
workspace.mkdir()
|
|
287
|
+
(workspace / "train.py").write_text("print('hello')")
|
|
288
|
+
job_toml = tmp_path / "job.toml"
|
|
289
|
+
job_toml.write_text(
|
|
290
|
+
'name = "test-job"\n'
|
|
291
|
+
'project = "test-project"\n'
|
|
292
|
+
'setup_command = "true"\n'
|
|
293
|
+
'run_command = "python train.py"\n'
|
|
294
|
+
'expires = "10m"\n'
|
|
295
|
+
"cpus = 1\n" + toml_body
|
|
296
|
+
)
|
|
297
|
+
|
|
298
|
+
with (
|
|
299
|
+
patch("trainwave_cli.cli.jobs.Api", return_value=mock_api),
|
|
300
|
+
patch("trainwave_cli.cli.jobs.Path.cwd", return_value=workspace),
|
|
301
|
+
):
|
|
302
|
+
result = await asyncio.to_thread(
|
|
303
|
+
runner.invoke, app, ["launch", "--config", str(job_toml)]
|
|
304
|
+
)
|
|
305
|
+
return result, mock_api
|
|
306
|
+
|
|
307
|
+
|
|
308
|
+
@pytest.mark.asyncio
|
|
309
|
+
async def test_launch_from_toml_without_env_vars(mock_job, tmp_path):
|
|
310
|
+
# Used to crash with "eval() arg 1 must be a string, bytes or code object".
|
|
311
|
+
result, mock_api = await _launch_from_toml(mock_job, tmp_path, "")
|
|
312
|
+
|
|
313
|
+
assert result.exit_code == 0, result.output
|
|
314
|
+
assert mock_api.create_job.call_args[0][0]["env_vars"] == {}
|
|
315
|
+
|
|
316
|
+
|
|
317
|
+
@pytest.mark.asyncio
|
|
318
|
+
async def test_launch_from_toml_env_vars_table(mock_job, tmp_path, monkeypatch):
|
|
319
|
+
monkeypatch.setenv("TW_TEST_TOKEN", "from-env")
|
|
320
|
+
result, mock_api = await _launch_from_toml(
|
|
321
|
+
mock_job,
|
|
322
|
+
tmp_path,
|
|
323
|
+
'[env_vars]\nSEED = "42"\nTOKEN = "${TW_TEST_TOKEN}"\n',
|
|
324
|
+
)
|
|
325
|
+
|
|
326
|
+
assert result.exit_code == 0, result.output
|
|
327
|
+
assert mock_api.create_job.call_args[0][0]["env_vars"] == {
|
|
328
|
+
"SEED": "42",
|
|
329
|
+
"TOKEN": "from-env",
|
|
330
|
+
}
|
|
331
|
+
|
|
332
|
+
|
|
333
|
+
@pytest.mark.parametrize(
|
|
334
|
+
("raw", "expected"),
|
|
335
|
+
[
|
|
336
|
+
(None, {}),
|
|
337
|
+
({}, {}),
|
|
338
|
+
("", {}),
|
|
339
|
+
("{}", {}),
|
|
340
|
+
('{"SEED": "42"}', {"SEED": "42"}),
|
|
341
|
+
({"SEED": 42}, {"SEED": "42"}),
|
|
342
|
+
],
|
|
343
|
+
)
|
|
344
|
+
def test_parse_env_vars(raw, expected):
|
|
345
|
+
assert _parse_env_vars(raw) == expected
|
|
346
|
+
|
|
347
|
+
|
|
348
|
+
@pytest.mark.parametrize("raw", ["__import__('os').getcwd()", "[1, 2]", "not python"])
|
|
349
|
+
def test_parse_env_vars_rejects_non_mappings_and_code(raw):
|
|
350
|
+
with pytest.raises((ValueError, SyntaxError, typer.BadParameter)):
|
|
351
|
+
_parse_env_vars(raw)
|