change train_on_completions to true

2026-02-19 06:02:16 +00:00 · 2026-02-19 06:02:16 +00:00 · e2b7b4b54c
commit e2b7b4b54c
parent 9f1caa1ceb
55 changed files with 55 additions and 55 deletions
--- a/studio/backend/assets/configs/model_defaults/default.yaml
+++ b/studio/backend/assets/configs/model_defaults/default.yaml
@ -14,7 +14,7 @@ training:
  weight_decay: 0.01
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml
+++ b/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.01
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: false
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml
@ -14,7 +14,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.01
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: true
  optim: "adamw_torch_fused"
  lr_scheduler_type: "cosine"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: true
  optim: "adamw_torch_fused"
  lr_scheduler_type: "cosine"
--- a/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.01
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: true
  optim: "adamw_torch"
  lr_scheduler_type: "cosine"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
+++ b/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "paged_adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml
@ -14,7 +14,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml
@ -14,7 +14,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.1
  random_seed: 3407
  packing: true
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4-14B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4-14B-Instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 42
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "paged_adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml
@ -15,7 +15,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml
@ -16,7 +16,7 @@ training:
  weight_decay: 0.001
  random_seed: 3407
  packing: false
-  train_on_completions: false
+  train_on_completions: true
  gradient_checkpointing: "unsloth"
  optim: "adamw_8bit"
  lr_scheduler_type: "linear"