ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

added model yaml config files

Browse source
This commit is contained in:
Roland Tannous 2026-02-02 18:22:14 +00:00
commit bfa03ebd3c
66 changed files with 3014 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

42
studio/backend/assets/configs/full_finetune.yaml Normal file
View file

@ -0,0 +1,42 @@
model: unsloth/Qwen2.5-0.5B
data:
dataset: tatsu-lab/alpaca
format_type: auto
training:
training_type: full
max_seq_length: 2048
load_in_4bit: false
output_dir: outputs
num_epochs: 1
learning_rate: 0.0002
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
lora:
lora_r: 64
lora_alpha: 16
lora_dropout: 0.0
target_modules: ""
vision_all_linear: false
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: unsloth-training
enable_tensorboard: false
tensorboard_dir: runs

42
studio/backend/assets/configs/lora_text.yaml Normal file
View file

@ -0,0 +1,42 @@
model: unsloth/Qwen2.5-0.5B
data:
dataset: tatsu-lab/alpaca
format_type: auto
training:
training_type: lora
max_seq_length: 2048
load_in_4bit: true
output_dir: outputs
num_epochs: 1
learning_rate: 0.0002
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
lora:
lora_r: 64
lora_alpha: 16
lora_dropout: 0.0
target_modules: "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj"
vision_all_linear: false
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: unsloth-training
enable_tensorboard: false
tensorboard_dir: runs

47
studio/backend/assets/configs/model_defaults/default.yaml Normal file
View file

@ -0,0 +1,47 @@
# Default model training parameters
# Used for models without specific configurations
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 5e-5
batch_size: 2
gradient_accumulation_steps: 4
warmup_ratio: 0.1
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/ERNIE-4.5-21B-A3B-PT
# Based on ERNIE_4_5_21B_A3B_PT-Conversational.ipynb
# Also applies to: unsloth/ERNIE-4.5-21B-A3B-PT
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 4
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/ERNIE-4.5-VL-28B-A3B-PT
# Based on ERNIE_4_5_VL_28B_A3B_PT_Vision.ipynb
# Also applies to: unsloth/ERNIE-4.5-VL-28B-A3B-PT
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for tiiuae/Falcon-H1-0.5B-Instruct
# Based on Falcon_H1_(0.5B)-Alpaca.ipynb
# Also applies to: tiiuae/Falcon-H1-0.5B-Instruct, unsloth/Falcon-H1-0.5B-Instruct
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 8
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: false
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.1
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

46
studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,46 @@
# Model defaults for unsloth/codegemma-7b-bnb-4bit
# Based on CodeGemma_(7B)-Conversational.ipynb
# Also applies to: unsloth/codegemma-7b, google/codegemma-7b
training:
max_seq_length: 4096
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/functiongemma-270m-it
# Based on FunctionGemma_(270M).ipynb
# Also applies to: unsloth/functiongemma-270m-it-unsloth-bnb-4bit, google/functiongemma-270m-it, unsloth/functiongemma-270m-it-unsloth-bnb-4bit
training:
max_seq_length: 4096
num_epochs: 4
learning_rate: 2e-4
batch_size: 4
gradient_accumulation_steps: 2
warmup_steps: 10
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 128
lora_alpha: 256
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

46
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,46 @@
# Model defaults for unsloth/gemma-2-27b-bnb-4bit
# Based on Gemma2_(9B)-Alpaca.ipynb (same defaults for larger models)
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/gemma-2-2b
# Based on Gemma2_(2B)-Alpaca.ipynb
# Also applies to: unsloth/gemma-2-2b-bnb-4bit, google/gemma-2-2b
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/gemma-3-270m-it
# Based on Gemma3_(270M).ipynb
# Also applies to: unsloth/gemma-3-270m-it-unsloth-bnb-4bit, google/gemma-3-270m-it, unsloth/gemma-3-270m-it-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 5e-5
batch_size: 4
gradient_accumulation_steps: 1
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 128
lora_alpha: 128
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/gemma-3-27b-it
# Based on Gemma3_(27B)_A100-Conversational.ipynb
# Also applies to: unsloth/gemma-3-27b-it-unsloth-bnb-4bit, google/gemma-3-27b-it, unsloth/gemma-3-27b-it-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 8
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/gemma-3-4b-it
# Based on Gemma3_(4B).ipynb
# Also applies to: unsloth/gemma-3-4b-it-unsloth-bnb-4bit, google/gemma-3-4b-it, unsloth/gemma-3-4b-it-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 8
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/gemma-3-4b-pt
# Based on Gemma3_(4B)-Vision.ipynb
# Also applies to: unsloth/gemma-3-4b-pt-unsloth-bnb-4bit, google/gemma-3-4b-pt, unsloth/gemma-3-4b-pt-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 2
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_ratio: 0.03
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: true
optim: "adamw_torch_fused"
lr_scheduler_type: "cosine"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/gemma-3n-E4B-it
# Based on Gemma3N_(4B)-Conversational.ipynb
# Also applies to: unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit, google/gemma-3n-E4B-it, unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit
training:
max_seq_length: 1024
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 8
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/gemma-3n-E4B
# Based on Gemma3N_(4B)-Vision.ipynb
# Also applies to: unsloth/gemma-3n-E4B-unsloth-bnb-4bit, google/gemma-3n-E4B
training:
max_seq_length: 2048
num_epochs: 2
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_ratio: 0.03
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: true
optim: "adamw_torch_fused"
lr_scheduler_type: "cosine"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/gpt-oss-120b
# Based on gpt-oss-(120B)_A100-Fine-tuning.ipynb
# Also applies to: openai/gpt-oss-120b, unsloth/gpt-oss-120b-unsloth-bnb-4bit
training:
max_seq_length: 4096
num_epochs: 4
learning_rate: 2e-4
batch_size: 4
gradient_accumulation_steps: 1
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/gpt-oss-20b
# Based on gpt-oss-(20B)-Fine-tuning.ipynb
# Also applies to: openai/gpt-oss-20b, unsloth/gpt-oss-20b-unsloth-bnb-4bit, unsloth/gpt-oss-20b-BF16
training:
max_seq_length: 1024
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

49
studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,49 @@
# Model defaults for unsloth/granite-4.0-350m
# Based on Granite4.0_350M.ipynb
# Also applies to: ibm-granite/granite-4.0-350m, unsloth/granite-4.0-350m-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
- "shared_mlp.input_linear"
- "shared_mlp.output_linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

49
studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml Normal file
View file

@ -0,0 +1,49 @@
# Model defaults for unsloth/granite-4.0-h-micro
# Based on Granite4.0.ipynb
# Also applies to: ibm-granite/granite-4.0-h-micro, unsloth/granite-4.0-h-micro-bnb-4bit, unsloth/granite-4.0-h-micro-unsloth-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
- "shared_mlp.input_linear"
- "shared_mlp.output_linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/Llama-3.2-11B-Vision-Instruct
# Based on Llama3.2_(11B)-Vision.ipynb
# Also applies to: unsloth/Llama-3.2-11B-Vision-Instruct-unsloth-bnb-4bit, meta-llama/Llama-3.2-11B-Vision-Instruct, unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Llama-3.2-1B-Instruct
# Based on Llama3.2_(1B)-RAFT.ipynb
# Also applies to: unsloth/Llama-3.2-1B-Instruct-unsloth-bnb-4bit, meta-llama/Llama-3.2-1B-Instruct, unsloth/Llama-3.2-1B-Instruct-bnb-4bit, RedHatAI/Llama-3.2-1B-Instruct-FP8, unsloth/Llama-3.2-1B-Instruct-FP8-Block, unsloth/Llama-3.2-1B-Instruct-FP8-Dynamic
training:
max_seq_length: 2048
num_epochs: 5
learning_rate: 2e-5
batch_size: 1
gradient_accumulation_steps: 8
warmup_steps: 0
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: true
optim: "adamw_torch"
lr_scheduler_type: "cosine"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Llama-3.2-3B-Instruct
# Based on Llama3.2_(1B_and_3B)-Conversational.ipynb
# Also applies to: unsloth/Llama-3.2-3B-Instruct-unsloth-bnb-4bit, meta-llama/Llama-3.2-3B-Instruct, unsloth/Llama-3.2-3B-Instruct-bnb-4bit, RedHatAI/Llama-3.2-3B-Instruct-FP8, unsloth/Llama-3.2-3B-Instruct-FP8-Block, unsloth/Llama-3.2-3B-Instruct-FP8-Dynamic
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Llama-3.3-70B-Instruct
# Based on Llama3.3_(70B)_A100-Conversational.ipynb
# Also applies to: unsloth/Llama-3.3-70B-Instruct-unsloth-bnb-4bit, meta-llama/Llama-3.3-70B-Instruct, unsloth/Llama-3.3-70B-Instruct-bnb-4bit, RedHatAI/Llama-3.3-70B-Instruct-FP8, unsloth/Llama-3.3-70B-Instruct-FP8-Block, unsloth/Llama-3.3-70B-Instruct-FP8-Dynamic
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Meta-Llama-3.1-70B-bnb-4bit
# Based on Llama3.1_(8B)-Alpaca.ipynb
# Also applies to: unsloth/Meta-Llama-3.1-8B-bnb-4bit, unsloth/Meta-Llama-3.1-8B-unsloth-bnb-4bit, meta-llama/Meta-Llama-3.1-8B, unsloth/Meta-Llama-3.1-8B, unsloth/Meta-Llama-3.1-70B, meta-llama/Meta-Llama-3.1-70B, unsloth/Meta-Llama-3.1-405B-bnb-4bit, meta-llama/Meta-Llama-3.1-405B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

48
studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,48 @@
# Model defaults for unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit
# Based on Llama3.1_(8B)-Inference.ipynb
# Also applies to: "unsloth/Meta-Llama-3.1-8B-Instruct-unsloth-bnb-4bit", "meta-llama/Meta-Llama-3.1-8B-Instruct", "unsloth/Meta-Llama-3.1-8B-Instruct","RedHatAI/Llama-3.1-8B-Instruct-FP8","unsloth/Llama-3.1-8B-Instruct-FP8-Block","unsloth/Llama-3.1-8B-Instruct-FP8-Dynamic"
training:
max_seq_length: 8192
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/llama-3-8b-Instruct-bnb-4bit
# Based on Llama3_(8B)-Conversational.ipynb
# Also applies to: unsloth/llama-3-8b-Instruct, meta-llama/Meta-Llama-3-8B-Instruct
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/llama-3-8b-bnb-4bit
# Based on Llama3_(8B)-Alpaca.ipynb
# Also applies to: unsloth/llama-3-8b, meta-llama/Meta-Llama-3-8B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

42
studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml Normal file
View file

@ -0,0 +1,42 @@
# Model defaults for unsloth/Llasa-3B
# Based on Llasa_TTS_(3B).ipynb and Llasa_TTS_(1B).ipynb
# Also applies to: HKUSTAudio/Llasa-1B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 5e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 128
lora_alpha: 128
lora_dropout: 0.0
target_modules:
- "q_proj"
- "v_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Magistral-Small-2509
# Based on Magistral_(24B)-Reasoning-Conversational.ipynb
# Also applies to: mistralai/Magistral-Small-2509, unsloth/Magistral-Small-2509-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Ministral-3-3B-Instruct-2512
# Based on Ministral_3_VL_(3B)_Vision.ipynb
# Also applies to: unsloth/Ministral-3-3B-Instruct-2512
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 4
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

48
studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,48 @@
# Model defaults for unsloth/Mistral-Nemo-Base-2407-bnb-4bit
# Based on Mistral_Nemo_(12B)-Alpaca.ipynb
# Also applies to: "unsloth/Mistral-Nemo-Base-2407", "mistralai/Mistral-Nemo-Base-2407", "unsloth/Mistral-Nemo-Instruct-2407-bnb-4bit", "unsloth/Mistral-Nemo-Instruct-2407", "mistralai/Mistral-Nemo-Instruct-2407",
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Mistral-Small-Instruct-2409
# Based on Mistral_Small_(22B)-Alpaca.ipynb
# Also applies to: unsloth/Mistral-Small-Instruct-2409-bnb-4bit, mistralai/Mistral-Small-Instruct-2409
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/Pixtral-12B-2409
# Based on Pixtral_(12B)-Vision.ipynb
# Also applies to: unsloth/Pixtral-12B-2409-unsloth-bnb-4bit, mistralai/Pixtral-12B-2409, unsloth/Pixtral-12B-2409-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "paged_adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 8
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: false
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

48
studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,48 @@
# Model defaults for unsloth/mistral-7b-instruct-v0.3-bnb-4bit
# Based on Mistral_v0.3_(7B)-Conversational.ipynb
# Also applies to: unsloth/mistral-7b-instruct-v0.3, mistralai/Mistral-7B-Instruct-v0.3
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/mistral-7b-v0.3-bnb-4bit
# Based on Mistral_v0.3_(7B)-Alpaca.ipynb
# Also applies to: "unsloth/mistral-7b-v0.3", "mistralai/Mistral-7B-v0.3",
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

42
studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml Normal file
View file

@ -0,0 +1,42 @@
# Model defaults for OuteAI/Llama-OuteTTS-1.0-1B
# Based on Oute_TTS_(1B).ipynb
# Also applies to: OuteAI/Llama-OuteTTS-1.0-1B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 128
lora_alpha: 128
lora_dropout: 0.0
target_modules:
- "q_proj"
- "v_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for Spark-TTS-0.5B/LLM
# Based on Spark_TTS_(0_5B).ipynb
# Also applies to: Spark-TTS-0.5B/LLM
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 128
lora_alpha: 128
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for sesame/csm-1b
# Based on Sesame_CSM_(1B)-TTS.ipynb
# Also applies to: sesame/csm-1b
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/LFM2-1.2B
# Based on Liquid_LFM2_(1.2B)-Conversational.ipynb
# Also applies to: unsloth/LFM2-1.2B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

49
studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml Normal file
View file

@ -0,0 +1,49 @@
# Model defaults for unsloth/Nemotron-3-Nano-30B-A3B
# Based on Nemotron-3-Nano-30B-A3B_A100.ipynb
# Also applies to: unsloth/Nemotron-3-Nano-30B-A3B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 4
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 8
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
- "in_proj"
- "out_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/PaddleOCR-VL
# Based on Paddle_OCR_(1B)_Vision.ipynb
# Also applies to: unsloth/PaddleOCR-VL
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 5e-5
batch_size: 4
gradient_accumulation_steps: 2
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 64
lora_alpha: 64
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

46
studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml Normal file
View file

@ -0,0 +1,46 @@
# Model defaults for answerdotai/ModernBERT-large
# Based on bert_classification.ipynb
training:
max_seq_length: 2048
num_epochs: 1
learning_rate: 5e-5
batch_size: 32
gradient_accumulation_steps: 1
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/orpheus-3b-0.1-ft
# Based on Orpheus_(3B)-TTS.ipynb
# Also applies to: unsloth/orpheus-3b-0.1-ft-unsloth-bnb-4bit, canopylabs/orpheus-3b-0.1-ft, unsloth/orpheus-3b-0.1-ft-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 64
lora_alpha: 64
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/tinyllama
# Based on TinyLlama_(1.1B)-Alpaca.ipynb
# Also applies to: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
training:
max_seq_length: 4096
num_epochs: 1
learning_rate: 2e-5
batch_size: 2
gradient_accumulation_steps: 4
warmup_ratio: 0.1
max_steps: 0
save_steps: 0
weight_decay: 0.1
random_seed: 3407
packing: true
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

42
studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml Normal file
View file

@ -0,0 +1,42 @@
# Model defaults for unsloth/whisper-large-v3
# Based on Whisper.ipynb
# Also applies to: unsloth/whisper-large-v3, openai/whisper-large-v3
training:
max_seq_length: 448
num_epochs: 4
learning_rate: 1e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 64
lora_alpha: 64
lora_dropout: 0.0
target_modules:
- "q_proj"
- "v_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

48
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml Normal file
View file

@ -0,0 +1,48 @@
# Model defaults for unsloth/Phi-3-medium-4k-instruct
# Based on Phi_3_Medium-Conversational.ipynb
# Also applies to: "unsloth/Phi-3-medium-4k-instruct-bnb-4bit", "microsoft/Phi-3-medium-4k-instruct",
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

48
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml Normal file
View file

@ -0,0 +1,48 @@
# Model defaults for unsloth/Phi-3.5-mini-instruct
# Based on Phi_3.5_Mini-Conversational.ipynb
# Also applies to: "unsloth/Phi-3.5-mini-instruct-bnb-4bit", "microsoft/Phi-3.5-mini-instruct"
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4-14B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Phi-4-14B-Instruct
# Based on Phi_4-Conversational.ipynb
# Also applies to: unsloth/Phi-4-14B-Instruct-unsloth-bnb-4bit, microsoft/Phi-4-14B-Instruct, unsloth/Phi-4-14B-Instruct-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Phi-4
# Based on Phi_4-Conversational.ipynb
# Also applies to: unsloth/phi-4-unsloth-bnb-4bit, microsoft/phi-4, unsloth/phi-4-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2-7B
# Based on Qwen2_(7B)-Alpaca.ipynb
# Also applies to: unsloth/Qwen2-7B-bnb-4bit, Qwen/Qwen2-7B
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/Qwen2-VL-7B-Instruct
# Based on Qwen2_VL_(7B)-Vision.ipynb
# Also applies to: unsloth/Qwen2-VL-7B-Instruct-unsloth-bnb-4bit, Qwen/Qwen2-VL-7B-Instruct, unsloth/Qwen2-VL-7B-Instruct-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2.5-1.5B-Instruct
# Based on nemo_gym_sudoku.ipynb
# Also applies to: unsloth/Qwen2.5-1.5B-Instruct-unsloth-bnb-4bit, Qwen/Qwen2.5-1.5B-Instruct, unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit
training:
max_seq_length: 4096
num_epochs: 4
learning_rate: 1e-5
batch_size: 1
gradient_accumulation_steps: 64
warmup_ratio: 0.1
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 42
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 4
lora_alpha: 8
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2.5-7B
# Based on Qwen2.5_(7B)-Alpaca.ipynb
# Also applies to: unsloth/Qwen2.5-7B-unsloth-bnb-4bit, Qwen/Qwen2.5-7B, unsloth/Qwen2.5-7B-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2.5-Coder-1.5B-Instruct
# Based on Qwen2.5_Coder_(1.5B)-Tool_Calling.ipynb
# Also applies to: unsloth/Qwen2.5-Coder-1.5B-Instruct-bnb-4bit, Qwen/Qwen2.5-Coder-1.5B-Instruct
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2.5-Coder-14B-Instruct
# Based on Qwen2.5_Coder_(14B)-Conversational.ipynb
# Also applies to: unsloth/Qwen2.5-Coder-14B-Instruct-bnb-4bit, Qwen/Qwen2.5-Coder-14B-Instruct
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "paged_adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen2.5-Coder-7B-Instruct-bnb-4bit
# Based on CodeForces-cot-Finetune_for_Reasoning_on_CodeForces.ipynb
# Also applies to: unsloth/Qwen2.5-Coder-7B-Instruct, Qwen/Qwen2.5-Coder-7B-Instruct
training:
max_seq_length: 32768
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/Qwen2.5-VL-7B-Instruct-bnb-4bit
# Based on Qwen2.5_VL_(7B)-Vision.ipynb
# Also applies to: unsloth/Qwen2.5-VL-7B-Instruct, Qwen/Qwen2.5-VL-7B-Instruct, unsloth/Qwen2.5-VL-7B-Instruct-unsloth-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-0.6B
# Based on Qwen3_(0_6B)-Phone_Deployment.ipynb
# Also applies to: unsloth/Qwen3-0.6B-unsloth-bnb-4bit, Qwen/Qwen3-0.6B, unsloth/Qwen3-0.6B-bnb-4bit, Qwen/Qwen3-0.6B-FP8, unsloth/Qwen3-0.6B-FP8
training:
max_seq_length: 1024
num_epochs: 4
learning_rate: 5e-5
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-14B-Base
# Based on Qwen3_(14B)-Alpaca.ipynb
# Also applies to: unsloth/Qwen3-14B-Base, Qwen/Qwen3-14B-Base, unsloth/Qwen3-14B-Base-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r:16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-14B
# Based on Qwen3_(14B).ipynb
# Also applies to: unsloth/Qwen3-14B-unsloth-bnb-4bit, Qwen/Qwen3-14B, unsloth/Qwen3-14B-bnb-4bit, Qwen/Qwen3-14B-FP8, unsloth/Qwen3-14B-FP8
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-32B
# Based on Qwen3_(32B)_A100-Reasoning-Conversational.ipynb
# Also applies to: unsloth/Qwen3-32B-unsloth-bnb-4bit, Qwen/Qwen3-32B, unsloth/Qwen3-32B-bnb-4bit, Qwen/Qwen3-32B-FP8, unsloth/Qwen3-32B-FP8
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-4B-Instruct-2507
# Based on Qwen3_(4B)-Instruct.ipynb
# Also applies to: unsloth/Qwen3-4B-Instruct-2507-unsloth-bnb-4bit, Qwen/Qwen3-4B-Instruct-2507, unsloth/Qwen3-4B-Instruct-2507-bnb-4bit, Qwen/Qwen3-4B-Instruct-2507-FP8, unsloth/Qwen3-4B-Instruct-2507-FP8
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

47
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml Normal file
View file

@ -0,0 +1,47 @@
# Model defaults for unsloth/Qwen3-4B-Thinking-2507
# Based on Qwen3_(4B)-Thinking.ipynb
# Also applies to: unsloth/Qwen3-4B-Thinking-2507-unsloth-bnb-4bit, Qwen/Qwen3-4B-Thinking-2507, unsloth/Qwen3-4B-Thinking-2507-bnb-4bit, Qwen/Qwen3-4B-Thinking-2507-FP8, unsloth/Qwen3-4B-Thinking-2507-FP8
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: true
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0
target_modules:
- "q_proj"
- "k_proj"
- "v_proj"
- "o_proj"
- "gate_proj"
- "up_proj"
- "down_proj"
use_rslora: false
use_loftq: false
finetune_vision_layers: false
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

41
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml Normal file
View file

@ -0,0 +1,41 @@
# Model defaults for unsloth/Qwen3-VL-8B-Instruct
# Based on Qwen3_VL_(8B)-Vision.ipynb
# Also applies to: Qwen/Qwen3-VL-8B-Instruct-FP8, unsloth/Qwen3-VL-8B-Instruct-FP8, unsloth/Qwen3-VL-8B-Instruct, Qwen/Qwen3-VL-8B-Instruct, unsloth/Qwen3-VL-8B-Instruct-bnb-4bit
training:
max_seq_length: 2048
num_epochs: 4
learning_rate: 2e-4
batch_size: 2
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.001
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
optim: "adamw_8bit"
lr_scheduler_type: "linear"
lora:
lora_r: 16
lora_alpha: 16
lora_dropout: 0.0
target_modules:
- "all-linear"
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: "llm-finetuning"
enable_tensorboard: false
tensorboard_dir: "runs"
log_frequency: 10

42
studio/backend/assets/configs/vision_lora.yaml Normal file
View file

@ -0,0 +1,42 @@
model: unsloth/Qwen2-VL-2B-Instruct-bnb-4bit
data:
dataset: philschmid/amazon-product-descriptions-vlm
format_type: auto
training:
training_type: lora
max_seq_length: 2048
load_in_4bit: true
output_dir: outputs
num_epochs: 1
learning_rate: 0.0002
batch_size: 1
gradient_accumulation_steps: 4
warmup_steps: 5
max_steps: 0
save_steps: 0
weight_decay: 0.01
random_seed: 3407
packing: false
train_on_completions: false
gradient_checkpointing: "unsloth"
lora:
lora_r: 64
lora_alpha: 16
lora_dropout: 0.0
target_modules: "" # vision uses vision_all_linear by default
vision_all_linear: true
use_rslora: false
use_loftq: false
finetune_vision_layers: true
finetune_language_layers: true
finetune_attention_modules: true
finetune_mlp_modules: true
logging:
enable_wandb: false
wandb_project: unsloth-training
enable_tensorboard: false
tensorboard_dir: runs