add support for https remote yamls (#1277)
Browse files- .mypy.ini +3 -0
- README.md +7 -0
- requirements-dev.txt +1 -0
- requirements.txt +1 -0
- src/axolotl/cli/__init__.py +53 -2
- src/axolotl/cli/preprocess.py +2 -1
- src/axolotl/cli/shard.py +2 -1
- src/axolotl/cli/train.py +2 -2
.mypy.ini
CHANGED
@@ -32,6 +32,9 @@ ignore_missing_imports = True
|
|
32 |
[mypy-bitsandbytes]
|
33 |
ignore_missing_imports = True
|
34 |
|
|
|
|
|
|
|
35 |
[mypy-datasets]
|
36 |
ignore_missing_imports = True
|
37 |
|
|
|
32 |
[mypy-bitsandbytes]
|
33 |
ignore_missing_imports = True
|
34 |
|
35 |
+
[mypy-requests]
|
36 |
+
ignore_missing_imports = True
|
37 |
+
|
38 |
[mypy-datasets]
|
39 |
ignore_missing_imports = True
|
40 |
|
README.md
CHANGED
@@ -121,6 +121,10 @@ accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
|
|
121 |
# gradio
|
122 |
accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
|
123 |
--lora_model_dir="./lora-out" --gradio
|
|
|
|
|
|
|
|
|
124 |
```
|
125 |
|
126 |
## Installation
|
@@ -988,6 +992,9 @@ Run
|
|
988 |
accelerate launch -m axolotl.cli.train your_config.yml
|
989 |
```
|
990 |
|
|
|
|
|
|
|
991 |
#### Preprocess dataset
|
992 |
|
993 |
You can optionally pre-tokenize dataset with the following before finetuning.
|
|
|
121 |
# gradio
|
122 |
accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
|
123 |
--lora_model_dir="./lora-out" --gradio
|
124 |
+
|
125 |
+
# remote yaml files - the yaml config can be hosted on a public URL
|
126 |
+
# Note: the yaml config must directly link to the **raw** yaml
|
127 |
+
accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/examples/openllama-3b/lora.yml
|
128 |
```
|
129 |
|
130 |
## Installation
|
|
|
992 |
accelerate launch -m axolotl.cli.train your_config.yml
|
993 |
```
|
994 |
|
995 |
+
> [!TIP]
|
996 |
+
> You can also reference a config file that is hosted on a public URL, for example `accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml`
|
997 |
+
|
998 |
#### Preprocess dataset
|
999 |
|
1000 |
You can optionally pre-tokenize dataset with the following before finetuning.
|
requirements-dev.txt
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
pre-commit
|
2 |
black
|
3 |
mypy
|
|
|
|
1 |
pre-commit
|
2 |
black
|
3 |
mypy
|
4 |
+
types-requests
|
requirements.txt
CHANGED
@@ -9,6 +9,7 @@ deepspeed>=0.13.1
|
|
9 |
addict
|
10 |
fire
|
11 |
PyYAML>=6.0
|
|
|
12 |
datasets>=2.15.0
|
13 |
flash-attn==2.3.3
|
14 |
sentencepiece
|
|
|
9 |
addict
|
10 |
fire
|
11 |
PyYAML>=6.0
|
12 |
+
requests
|
13 |
datasets>=2.15.0
|
14 |
flash-attn==2.3.3
|
15 |
sentencepiece
|
src/axolotl/cli/__init__.py
CHANGED
@@ -1,16 +1,20 @@
|
|
1 |
"""Prepare and train a model on a dataset. Can also infer from a model or merge lora"""
|
2 |
|
3 |
import importlib
|
|
|
4 |
import logging
|
5 |
import math
|
6 |
import os
|
7 |
import random
|
8 |
import sys
|
|
|
9 |
from pathlib import Path
|
10 |
from threading import Thread
|
11 |
from typing import Any, Dict, List, Optional, Union
|
|
|
12 |
|
13 |
import gradio as gr
|
|
|
14 |
import torch
|
15 |
import yaml
|
16 |
|
@@ -59,6 +63,52 @@ def print_axolotl_text_art(suffix=None):
|
|
59 |
print(ascii_art)
|
60 |
|
61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
def get_multi_line_input() -> Optional[str]:
|
63 |
print("Give me an instruction (Ctrl + D to submit): ")
|
64 |
instruction = ""
|
@@ -270,9 +320,10 @@ def check_not_in(list1: List[str], list2: Union[Dict[str, Any], List[str]]) -> b
|
|
270 |
return not any(el in list2 for el in list1)
|
271 |
|
272 |
|
273 |
-
def load_cfg(config: Path = Path("examples/"), **kwargs):
|
|
|
274 |
if Path(config).is_dir():
|
275 |
-
config = choose_config(config)
|
276 |
|
277 |
# load the config from the yaml file
|
278 |
with open(config, encoding="utf-8") as file:
|
|
|
1 |
"""Prepare and train a model on a dataset. Can also infer from a model or merge lora"""
|
2 |
|
3 |
import importlib
|
4 |
+
import json
|
5 |
import logging
|
6 |
import math
|
7 |
import os
|
8 |
import random
|
9 |
import sys
|
10 |
+
import tempfile
|
11 |
from pathlib import Path
|
12 |
from threading import Thread
|
13 |
from typing import Any, Dict, List, Optional, Union
|
14 |
+
from urllib.parse import urlparse
|
15 |
|
16 |
import gradio as gr
|
17 |
+
import requests
|
18 |
import torch
|
19 |
import yaml
|
20 |
|
|
|
63 |
print(ascii_art)
|
64 |
|
65 |
|
66 |
+
def check_remote_config(config: Union[str, Path]):
|
67 |
+
# Check if the config is a valid HTTPS URL to a .yml or .yaml file
|
68 |
+
if not (isinstance(config, str) and config.startswith("https://")):
|
69 |
+
return config # Return the original value if it's not a valid URL
|
70 |
+
|
71 |
+
filename = os.path.basename(urlparse(config).path)
|
72 |
+
temp_dir = tempfile.mkdtemp()
|
73 |
+
|
74 |
+
try:
|
75 |
+
response = requests.get(config, timeout=30)
|
76 |
+
response.raise_for_status() # Check for HTTP errors
|
77 |
+
|
78 |
+
content = response.content
|
79 |
+
try:
|
80 |
+
# Try parsing as JSON first to catch cases where JSON content is mistakenly considered YAML
|
81 |
+
json.loads(content)
|
82 |
+
# Log a warning but do not raise an error; JSON is technically valid YAML - this can happen when you forget to point to a raw github link
|
83 |
+
LOG.warning(
|
84 |
+
f"Warning: The content of the file at {config} is JSON, which is technically valid YAML but might not be intended."
|
85 |
+
)
|
86 |
+
except json.JSONDecodeError:
|
87 |
+
# If it's not valid JSON, verify it's valid YAML
|
88 |
+
try:
|
89 |
+
yaml.safe_load(content)
|
90 |
+
except yaml.YAMLError as err:
|
91 |
+
raise ValueError(
|
92 |
+
f"Failed to parse the content at {config} as YAML: {err}"
|
93 |
+
) from err
|
94 |
+
|
95 |
+
# Write the content to a file if it's valid YAML (or JSON treated as YAML)
|
96 |
+
output_path = Path(temp_dir) / filename
|
97 |
+
with open(output_path, "wb") as file:
|
98 |
+
file.write(content)
|
99 |
+
LOG.info(
|
100 |
+
f"Using the following config obtained from {config}:\n\n{content.decode('utf-8')}\n"
|
101 |
+
)
|
102 |
+
return output_path
|
103 |
+
|
104 |
+
except requests.RequestException as err:
|
105 |
+
# This catches all requests-related exceptions including HTTPError
|
106 |
+
raise RuntimeError(f"Failed to download {config}: {err}") from err
|
107 |
+
except Exception as err:
|
108 |
+
# Catch-all for any other exceptions
|
109 |
+
raise err
|
110 |
+
|
111 |
+
|
112 |
def get_multi_line_input() -> Optional[str]:
|
113 |
print("Give me an instruction (Ctrl + D to submit): ")
|
114 |
instruction = ""
|
|
|
320 |
return not any(el in list2 for el in list1)
|
321 |
|
322 |
|
323 |
+
def load_cfg(config: Union[str, Path] = Path("examples/"), **kwargs):
|
324 |
+
config = check_remote_config(config)
|
325 |
if Path(config).is_dir():
|
326 |
+
config = choose_config(Path(config))
|
327 |
|
328 |
# load the config from the yaml file
|
329 |
with open(config, encoding="utf-8") as file:
|
src/axolotl/cli/preprocess.py
CHANGED
@@ -3,6 +3,7 @@ CLI to run training on a model
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
|
|
6 |
|
7 |
import fire
|
8 |
import transformers
|
@@ -23,7 +24,7 @@ from axolotl.prompt_strategies.sharegpt import register_chatml_template
|
|
23 |
LOG = logging.getLogger("axolotl.cli.preprocess")
|
24 |
|
25 |
|
26 |
-
def do_cli(config: Path = Path("examples/"), **kwargs):
|
27 |
# pylint: disable=duplicate-code
|
28 |
print_axolotl_text_art()
|
29 |
parsed_cfg = load_cfg(config, **kwargs)
|
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
6 |
+
from typing import Union
|
7 |
|
8 |
import fire
|
9 |
import transformers
|
|
|
24 |
LOG = logging.getLogger("axolotl.cli.preprocess")
|
25 |
|
26 |
|
27 |
+
def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
|
28 |
# pylint: disable=duplicate-code
|
29 |
print_axolotl_text_art()
|
30 |
parsed_cfg = load_cfg(config, **kwargs)
|
src/axolotl/cli/shard.py
CHANGED
@@ -3,6 +3,7 @@ CLI to shard a trained model into 10GiB chunks
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
|
|
6 |
|
7 |
import fire
|
8 |
import transformers
|
@@ -25,7 +26,7 @@ def shard(
|
|
25 |
model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
|
26 |
|
27 |
|
28 |
-
def do_cli(config: Path = Path("examples/"), **kwargs):
|
29 |
# pylint: disable=duplicate-code
|
30 |
print_axolotl_text_art()
|
31 |
parsed_cfg = load_cfg(config, **kwargs)
|
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
6 |
+
from typing import Union
|
7 |
|
8 |
import fire
|
9 |
import transformers
|
|
|
26 |
model.save_pretrained(cfg.output_dir, safe_serialization=safe_serialization)
|
27 |
|
28 |
|
29 |
+
def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
|
30 |
# pylint: disable=duplicate-code
|
31 |
print_axolotl_text_art()
|
32 |
parsed_cfg = load_cfg(config, **kwargs)
|
src/axolotl/cli/train.py
CHANGED
@@ -3,7 +3,7 @@ CLI to run training on a model
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
6 |
-
from typing import Tuple
|
7 |
|
8 |
import fire
|
9 |
from transformers.hf_argparser import HfArgumentParser
|
@@ -25,7 +25,7 @@ from axolotl.train import train
|
|
25 |
LOG = logging.getLogger("axolotl.cli.train")
|
26 |
|
27 |
|
28 |
-
def do_cli(config: Path = Path("examples/"), **kwargs):
|
29 |
# pylint: disable=duplicate-code
|
30 |
parsed_cfg = load_cfg(config, **kwargs)
|
31 |
parser = HfArgumentParser((TrainerCliArgs))
|
|
|
3 |
"""
|
4 |
import logging
|
5 |
from pathlib import Path
|
6 |
+
from typing import Tuple, Union
|
7 |
|
8 |
import fire
|
9 |
from transformers.hf_argparser import HfArgumentParser
|
|
|
25 |
LOG = logging.getLogger("axolotl.cli.train")
|
26 |
|
27 |
|
28 |
+
def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
|
29 |
# pylint: disable=duplicate-code
|
30 |
parsed_cfg = load_cfg(config, **kwargs)
|
31 |
parser = HfArgumentParser((TrainerCliArgs))
|