Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[FEATURE] Default model list can be configured with ENV vars #870

Merged
merged 4 commits into from
Sep 26, 2024
Merged

[FEATURE] Default model list can be configured with ENV vars #870

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
d0739be
move model list to default_cfg
pascal-pfeiffer Sep 26, 2024
4d8985d
H2O_LLM_STUDIO_DEFAULT_LM_MODELS & H2O_LLM_STUDIO_DEFAULT_S2S_MODELS
pascal-pfeiffer Sep 26, 2024
4d4f42a
format
pascal-pfeiffer Sep 26, 2024
fc76232
Merge branch 'main' into pp/override_default_models
pascal-pfeiffer Sep 26, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 43 additions & 0 deletions llm_studio/app_utils/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,47 @@ def get_size(x):
url = f"http://{host}:{port}/"


if os.getenv("H2O_LLM_STUDIO_DEFAULT_LM_MODELS"):
default_causal_language_models = [
mdl.strip() for mdl in os.getenv("H2O_LLM_STUDIO_DEFAULT_LM_MODELS").split(",")
]
else:
default_causal_language_models = [
"h2oai/h2o-danube3-500m-base",
"h2oai/h2o-danube3-500m-chat",
"h2oai/h2o-danube3-4b-base",
"h2oai/h2o-danube3-4b-chat",
"h2oai/h2o-danube2-1.8b-base",
"h2oai/h2o-danube2-1.8b-chat",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Meta-Llama-3.1-8B-Instruct",
"meta-llama/Meta-Llama-3.1-70B-Instruct",
"mistralai/Mistral-7B-v0.3",
"mistralai/Mistral-7B-Instruct-v0.2",
"google/gemma-2-2b-it",
"google/gemma-2-9b-it",
"microsoft/Phi-3-mini-4k-instruct",
"microsoft/Phi-3-medium-4k-instruct",
"Qwen/Qwen2-7B-Instruct",
"Qwen/Qwen2-72B-Instruct",
]

if os.getenv("H2O_LLM_STUDIO_DEFAULT_S2S_MODELS"):
default_sequence_to_sequence_models = [
mdl.strip() for mdl in os.getenv("H2O_LLM_STUDIO_DEFAULT_S2S_MODELS").split(",")
]
else:
default_sequence_to_sequence_models = [
"t5-small",
"t5-base",
"t5-large",
"google/flan-t5-small",
"google/flan-t5-base",
"google/flan-t5-large",
"google/flan-ul2",
]

default_cfg = {
"url": url,
"name": "H2O LLM Studio",
Expand All @@ -67,6 +108,8 @@ def get_size(x):
"text_sequence_to_sequence_modeling_config",
"text_dpo_modeling_config",
],
"default_causal_language_models": default_causal_language_models,
"default_sequence_to_sequence_models": default_sequence_to_sequence_models,
"problem_categories": ["text"],
"dataset_keys": [
"train_dataframe",
Expand Down
28 changes: 7 additions & 21 deletions llm_studio/python_configs/text_causal_classification_modeling_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@

import llm_studio.src.datasets.text_causal_classification_ds
import llm_studio.src.plots.text_causal_classification_modeling_plots
from llm_studio.app_utils.config import default_cfg
from llm_studio.python_configs.base import DefaultConfig, DefaultConfigProblemBase
from llm_studio.python_configs.text_causal_language_modeling_config import (
ConfigNLPAugmentation,
Expand Down Expand Up @@ -155,7 +156,11 @@ class ConfigNLPCausalClassificationLogging(ConfigNLPCausalLMLogging):
class ConfigProblemBase(DefaultConfigProblemBase):
output_directory: str = f"output/{os.path.basename(__file__).split('.')[0]}"
experiment_name: str = field(default_factory=generate_experiment_name)
llm_backbone: str = "h2oai/h2o-danube3-500m-chat"
llm_backbone: str = (
"h2oai/h2o-danube3-500m-chat"
if "h2oai/h2o-danube3-500m-chat" in default_cfg.default_causal_language_models
else default_cfg.default_causal_language_models[0]
)

dataset: ConfigNLPCausalClassificationDataset = field(
default_factory=ConfigNLPCausalClassificationDataset
Expand Down Expand Up @@ -188,26 +193,7 @@ def __post_init__(self):
self._visibility["output_directory"] = -1

self._possible_values["llm_backbone"] = possible_values.String(
values=(
"h2oai/h2o-danube3-500m-base",
"h2oai/h2o-danube3-500m-chat",
"h2oai/h2o-danube3-4b-base",
"h2oai/h2o-danube3-4b-chat",
"h2oai/h2o-danube2-1.8b-base",
"h2oai/h2o-danube2-1.8b-chat",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Meta-Llama-3.1-8B-Instruct",
"meta-llama/Meta-Llama-3.1-70B-Instruct",
"mistralai/Mistral-7B-v0.3",
"mistralai/Mistral-7B-Instruct-v0.2",
"google/gemma-2-2b-it",
"google/gemma-2-9b-it",
"microsoft/Phi-3-mini-4k-instruct",
"microsoft/Phi-3-medium-4k-instruct",
"Qwen/Qwen2-7B-Instruct",
"Qwen/Qwen2-72B-Instruct",
),
values=default_cfg.default_causal_language_models,
allow_custom=True,
)

Expand Down
28 changes: 7 additions & 21 deletions llm_studio/python_configs/text_causal_language_modeling_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@
import torch

import llm_studio.src.datasets.text_causal_language_modeling_ds
from llm_studio.app_utils.config import default_cfg
from llm_studio.python_configs.base import DefaultConfig, DefaultConfigProblemBase
from llm_studio.src import possible_values
from llm_studio.src.augmentations.nlp_aug import BaseNLPAug
Expand Down Expand Up @@ -613,7 +614,11 @@ def __post_init__(self):
class ConfigProblemBase(DefaultConfigProblemBase):
output_directory: str = f"output/{os.path.basename(__file__).split('.')[0]}"
experiment_name: str = field(default_factory=generate_experiment_name)
llm_backbone: str = "h2oai/h2o-danube3-500m-base"
llm_backbone: str = (
"h2oai/h2o-danube3-500m-base"
if "h2oai/h2o-danube3-500m-base" in default_cfg.default_causal_language_models
else default_cfg.default_causal_language_models[0]
)

dataset: ConfigNLPCausalLMDataset = field(default_factory=ConfigNLPCausalLMDataset)
tokenizer: ConfigNLPCausalLMTokenizer = field(
Expand All @@ -640,26 +645,7 @@ def __post_init__(self):
self._visibility["output_directory"] = -1

self._possible_values["llm_backbone"] = possible_values.String(
values=(
"h2oai/h2o-danube3-500m-base",
"h2oai/h2o-danube3-500m-chat",
"h2oai/h2o-danube3-4b-base",
"h2oai/h2o-danube3-4b-chat",
"h2oai/h2o-danube2-1.8b-base",
"h2oai/h2o-danube2-1.8b-chat",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Meta-Llama-3.1-8B-Instruct",
"meta-llama/Meta-Llama-3.1-70B-Instruct",
"mistralai/Mistral-7B-v0.3",
"mistralai/Mistral-7B-Instruct-v0.2",
"google/gemma-2-2b-it",
"google/gemma-2-9b-it",
"microsoft/Phi-3-mini-4k-instruct",
"microsoft/Phi-3-medium-4k-instruct",
"Qwen/Qwen2-7B-Instruct",
"Qwen/Qwen2-72B-Instruct",
),
values=default_cfg.default_causal_language_models,
allow_custom=True,
)

Expand Down
28 changes: 7 additions & 21 deletions llm_studio/python_configs/text_causal_regression_modeling_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@

import llm_studio.src.datasets.text_causal_regression_ds
import llm_studio.src.plots.text_causal_classification_modeling_plots
from llm_studio.app_utils.config import default_cfg
from llm_studio.python_configs.base import DefaultConfig, DefaultConfigProblemBase
from llm_studio.python_configs.text_causal_classification_modeling_config import (
ConfigNLPCausalClassificationAugmentation as ConfigNLPCausalRegressionAugmentation,
Expand Down Expand Up @@ -109,7 +110,11 @@ def __post_init__(self):
class ConfigProblemBase(DefaultConfigProblemBase):
output_directory: str = f"output/{os.path.basename(__file__).split('.')[0]}"
experiment_name: str = field(default_factory=generate_experiment_name)
llm_backbone: str = "h2oai/h2o-danube3-500m-chat"
llm_backbone: str = (
"h2oai/h2o-danube3-500m-chat"
if "h2oai/h2o-danube3-500m-chat" in default_cfg.default_causal_language_models
else default_cfg.default_causal_language_models[0]
)

dataset: ConfigNLPCausalRegressionDataset = field(
default_factory=ConfigNLPCausalRegressionDataset
Expand Down Expand Up @@ -142,26 +147,7 @@ def __post_init__(self):
self._visibility["output_directory"] = -1

self._possible_values["llm_backbone"] = possible_values.String(
values=(
"h2oai/h2o-danube3-500m-base",
"h2oai/h2o-danube3-500m-chat",
"h2oai/h2o-danube3-4b-base",
"h2oai/h2o-danube3-4b-chat",
"h2oai/h2o-danube2-1.8b-base",
"h2oai/h2o-danube2-1.8b-chat",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Meta-Llama-3.1-8B-Instruct",
"meta-llama/Meta-Llama-3.1-70B-Instruct",
"mistralai/Mistral-7B-v0.3",
"mistralai/Mistral-7B-Instruct-v0.2",
"google/gemma-2-2b-it",
"google/gemma-2-9b-it",
"microsoft/Phi-3-mini-4k-instruct",
"microsoft/Phi-3-medium-4k-instruct",
"Qwen/Qwen2-7B-Instruct",
"Qwen/Qwen2-72B-Instruct",
),
values=default_cfg.default_causal_language_models,
allow_custom=True,
)

Expand Down
28 changes: 7 additions & 21 deletions llm_studio/python_configs/text_dpo_modeling_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
from typing import Any

import llm_studio.src.datasets.text_dpo_modeling_ds
from llm_studio.app_utils.config import default_cfg
from llm_studio.python_configs.base import DefaultConfigProblemBase
from llm_studio.python_configs.text_causal_language_modeling_config import (
ConfigNLPAugmentation,
Expand Down Expand Up @@ -108,7 +109,11 @@ class ConfigProblemBase(DefaultConfigProblemBase):
output_directory: str = f"output/{os.path.basename(__file__).split('.')[0]}"
experiment_name: str = field(default_factory=generate_experiment_name)

llm_backbone: str = "h2oai/h2o-danube3-500m-chat"
llm_backbone: str = (
"h2oai/h2o-danube3-500m-chat"
if "h2oai/h2o-danube3-500m-chat" in default_cfg.default_causal_language_models
else default_cfg.default_causal_language_models[0]
)

dataset: ConfigDPODataset = field(default_factory=ConfigDPODataset)
tokenizer: ConfigNLPCausalLMTokenizer = field(
Expand All @@ -129,25 +134,6 @@ def __post_init__(self):
super().__post_init__()
self._visibility["output_directory"] = -1
self._possible_values["llm_backbone"] = possible_values.String(
values=(
"h2oai/h2o-danube3-500m-base",
"h2oai/h2o-danube3-500m-chat",
"h2oai/h2o-danube3-4b-base",
"h2oai/h2o-danube3-4b-chat",
"h2oai/h2o-danube2-1.8b-base",
"h2oai/h2o-danube2-1.8b-chat",
"meta-llama/Llama-3.2-1B-Instruct",
"meta-llama/Llama-3.2-3B-Instruct",
"meta-llama/Meta-Llama-3.1-8B-Instruct",
"meta-llama/Meta-Llama-3.1-70B-Instruct",
"mistralai/Mistral-7B-v0.3",
"mistralai/Mistral-7B-Instruct-v0.2",
"google/gemma-2-2b-it",
"google/gemma-2-9b-it",
"microsoft/Phi-3-mini-4k-instruct",
"microsoft/Phi-3-medium-4k-instruct",
"Qwen/Qwen2-7B-Instruct",
"Qwen/Qwen2-72B-Instruct",
),
values=default_cfg.default_causal_language_models,
allow_custom=True,
)
17 changes: 7 additions & 10 deletions llm_studio/python_configs/text_sequence_to_sequence_modeling_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
from dataclasses import dataclass, field
from typing import Any, Dict, List

from llm_studio.app_utils.config import default_cfg
from llm_studio.python_configs.base import DefaultConfigProblemBase
from llm_studio.python_configs.text_causal_language_modeling_config import (
ConfigNLPAugmentation,
Expand Down Expand Up @@ -72,7 +73,11 @@ def __post_init__(self):
class ConfigProblemBase(DefaultConfigProblemBase):
output_directory: str = f"output/{os.path.basename(__file__).split('.')[0]}"
experiment_name: str = field(default_factory=generate_experiment_name)
llm_backbone: str = "t5-small"
llm_backbone: str = (
"t5-small"
if "t5-small" in default_cfg.default_sequence_to_sequence_models
else default_cfg.default_sequence_to_sequence_models[0]
)

dataset: ConfigNLPSeq2SeqDataset = field(default_factory=ConfigNLPSeq2SeqDataset)
tokenizer: ConfigNLPCausalLMTokenizer = field(
Expand All @@ -99,15 +104,7 @@ def __post_init__(self):
self._visibility["output_directory"] = -1

self._possible_values["llm_backbone"] = possible_values.String(
values=(
"t5-small",
"t5-base",
"t5-large",
"google/flan-t5-small",
"google/flan-t5-base",
"google/flan-t5-large",
"google/flan-ul2",
),
values=default_cfg.default_sequence_to_sequence_models,
allow_custom=True,
)

Expand Down