Merge pull request #338 from unslothai/fix/trust-code

Exposed trust_remote_code through the UI
2026-03-09 17:19:56 +04:00 · 2026-03-09 17:19:56 +04:00 · 87c8d7b3da
commit 87c8d7b3da
parent e1b6798a4e 91dd7fc762
92 changed files with 291 additions and 21 deletions
--- a/studio/backend/assets/configs/model_defaults/default.yaml
+++ b/studio/backend/assets/configs/model_defaults/default.yaml
@ -2,6 +2,7 @@
 # Used for models without specific configurations

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -47,6 +48,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.7
  top_p: 0.95
  top_k: -1
--- a/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml
+++ b/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml
@ -3,6 +3,7 @@
 # Also applies to: unsloth/ERNIE-4.5-21B-A3B-PT

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: true
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -48,6 +49,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: true
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # Also applies to: tiiuae/Falcon-H1-0.5B-Instruct, unsloth/Falcon-H1-0.5B-Instruct

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from Ollama

 training:
+  trust_remote_code: false
  max_seq_length: 4096
  # num_epochs: 4
  num_epochs: 0
@ -44,5 +45,6 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0
  top_p: 0.9
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 4096
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -2,6 +2,7 @@
 # Based on Gemma2_(9B)-Alpaca.ipynb (same defaults for larger models)

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -41,3 +42,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/gemma-2-2b-bnb-4bit, google/gemma-2-2b

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 2
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 1024
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
 audio_input: true

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 2
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
 audio_input: true

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_k: 64
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 4096
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_p: 1.0
  top_k: 0
--- a/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 1024
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.0
  top_p: 1.0
  top_k: 0
--- a/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -46,6 +47,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.0
  top_p: 1.0
  top_k: 0
--- a/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml
+++ b/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -46,6 +47,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.0
  top_p: 1.0
  top_k: 0
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # 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:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 5
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # 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:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # 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:
+  trust_remote_code: false
  max_seq_length: 8192
  # num_epochs: 4
  num_epochs: 0
@ -42,4 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/llama-3-8b-Instruct, meta-llama/Meta-Llama-3-8B-Instruct

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/llama-3-8b, meta-llama/Meta-Llama-3-8B

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- a/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
+++ b/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -39,6 +40,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.2
  top_p: 1.2

--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -48,6 +49,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.7
  min_p: 0.01
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -48,6 +49,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.15
  top_p: default

--- 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
@ -3,6 +3,7 @@
 # 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:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,4 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/Mistral-Small-Instruct-2409-bnb-4bit, mistralai/Mistral-Small-Instruct-2409

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/mistral-7b-instruct-v0.3, mistralai/Mistral-7B-Instruct-v0.3

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,4 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- 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
@ -2,6 +2,7 @@
 # Based on Mistral_v0.3_(7B)-Alpaca.ipynb
 # Also applies to: "unsloth/mistral-7b-v0.3", "mistralai/Mistral-7B-v0.3",
 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -41,4 +42,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- 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
@ -6,6 +6,7 @@
 audio_type: dac

 training:
+  trust_remote_code: false
  eval_steps: 0
  max_seq_length: 2048
  # num_epochs: 4
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.4
  top_k: 40
  top_p: 0.9
--- 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
@ -6,6 +6,7 @@
 audio_type: bicodec

 training:
+  trust_remote_code: false
  eval_steps: 0
  max_seq_length: 2048
  # num_epochs: 4
@ -47,6 +48,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.8
  top_k: 50
  top_p: 1.0
--- a/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
@ -5,6 +5,7 @@
 audio_type: csm

 training:
+  trust_remote_code: false
  eval_steps: 0
  max_seq_length: 2048
  # num_epochs: 4
@ -45,3 +46,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_GLM-4.7-Flash.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_GLM-4.7-Flash.yaml
@ -3,6 +3,7 @@
 # Also applies to: unsloth/GLM-4.7-Flash-unsloth-bnb-4bit, unsloth/GLM-4.7-Flash-bnb-4bit, THUDM/GLM-4.7-Flash

 training:
+  trust_remote_code: true
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: true
  temperature: 0.7
  top_p: 0.8
  top_k: 20
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -38,6 +39,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.3
  min_p: 0.15

--- a/studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: true
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -46,6 +47,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: true
  temperature: 1.0
  top_p: 1.0

--- a/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: true
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -48,6 +49,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: true
  temperature: 1.5
  min_p: 0.1

--- 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
@ -2,6 +2,7 @@
 # Based on bert_classification.ipynb

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 1
  num_epochs: 0
@ -41,3 +42,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -6,6 +6,7 @@
 audio_type: snac

 training:
+  trust_remote_code: false
  eval_steps: 0
  max_seq_length: 2048
  # num_epochs: 4
@ -47,6 +48,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_p: 0.95

--- 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
@ -3,6 +3,7 @@
 # Also applies to: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T

 training:
+  trust_remote_code: false
  max_seq_length: 4096
  # num_epochs: 1
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -6,6 +6,7 @@ audio_type: whisper
 audio_input: true

 training:
+  trust_remote_code: false
  eval_steps: 5
  max_seq_length: 448
  # num_epochs: 4
@ -41,3 +42,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: "unsloth/Phi-3-medium-4k-instruct-bnb-4bit", "microsoft/Phi-3-medium-4k-instruct",

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,4 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: "unsloth/Phi-3.5-mini-instruct-bnb-4bit", "microsoft/Phi-3.5-mini-instruct"

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,4 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

-
+inference:
+  trust_remote_code: false
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.8
  top_p: 0.95

--- a/studio/backend/assets/configs/model_defaults/qwen/imdatta0_tiny_qwen3_moe_2.8B_0.7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/imdatta0_tiny_qwen3_moe_2.8B_0.7B.yaml
@ -4,6 +4,7 @@
 # MoE model - includes gate_up_proj for MoE layers

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -45,6 +46,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
@ -3,6 +3,7 @@
 # Also applies to: unsloth/Qwen2-7B-bnb-4bit, Qwen/Qwen2-7B

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # 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:
+  trust_remote_code: false
  max_seq_length: 4096
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/Qwen2.5-7B-unsloth-bnb-4bit, Qwen/Qwen2.5-7B, unsloth/Qwen2.5-7B-bnb-4bit

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/Qwen2.5-Coder-1.5B-Instruct-bnb-4bit, Qwen/Qwen2.5-Coder-1.5B-Instruct

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -3,6 +3,7 @@
 # Also applies to: unsloth/Qwen2.5-Coder-7B-Instruct, Qwen/Qwen2.5-Coder-7B-Instruct

 training:
+  trust_remote_code: false
  max_seq_length: 32768
  # num_epochs: 4
  num_epochs: 0
@ -42,3 +43,5 @@ logging:
  tensorboard_dir: "runs"
  log_frequency: 10

+inference:
+  trust_remote_code: false
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth notebook

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 1.5
  min_p: 0.1

--- 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
@ -4,6 +4,7 @@
 # added inference parameters from Ollama

 training:
+  trust_remote_code: false
  max_seq_length: 1024
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from Ollama

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
@ -4,6 +4,7 @@
 # added inference parameters from Ollama

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-30B-A3B-Instruct-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-30B-A3B-Instruct-2507.yaml
@ -4,6 +4,7 @@
 # MoE model - includes gate_up_proj for MoE layers

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -45,6 +46,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
@ -4,6 +4,7 @@
 # added inference parameters from Ollama

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_k: 20
  top_p: 0.95
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.7
  top_p: 0.80
  top_k: 20
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -44,6 +45,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.6
  top_p: 0.95
  top_k: 20
--- 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
@ -4,6 +4,7 @@
 # added inference parameters from unsloth guides

 training:
+  trust_remote_code: false
  max_seq_length: 2048
  # num_epochs: 4
  num_epochs: 0
@ -42,6 +43,7 @@ logging:
  log_frequency: 10

 inference:
+  trust_remote_code: false
  temperature: 0.7
  top_p: 0.8
  top_k: 20
--- a/studio/backend/core/export/export.py
+++ b/studio/backend/core/export/export.py
@ -138,7 +138,8 @@ class ExportBackend:
    def load_checkpoint(self,
                       checkpoint_path: str,
                       max_seq_length: int = 2048,
-                       load_in_4bit: bool = True) -> Tuple[bool, str]:
+                       load_in_4bit: bool = True,
+                       trust_remote_code: bool = False) -> Tuple[bool, str]:
        """
        Load a checkpoint for export.

@ -178,6 +179,7 @@ class ExportBackend:
                    dtype=None,
                    auto_model=CsmForConditionalGeneration,
                    load_in_4bit=False,
+                    trust_remote_code=trust_remote_code,
                )

            elif self._audio_type == 'whisper':
@ -189,6 +191,7 @@ class ExportBackend:
                    dtype=None,
                    load_in_4bit=False,
                    auto_model=WhisperForConditionalGeneration,
+                    trust_remote_code=trust_remote_code,
                )

            elif self._audio_type == 'snac':
@ -198,6 +201,7 @@ class ExportBackend:
                    max_seq_length=max_seq_length,
                    dtype=None,
                    load_in_4bit=load_in_4bit,
+                    trust_remote_code=trust_remote_code,
                )

            elif self._audio_type == 'bicodec':
@ -208,6 +212,7 @@ class ExportBackend:
                    max_seq_length=max_seq_length,
                    dtype=torch.float32,
                    load_in_4bit=False,
+                    trust_remote_code=trust_remote_code,
                )

            elif self._audio_type == 'dac':
@ -217,6 +222,7 @@ class ExportBackend:
                    model_name=checkpoint_path,
                    max_seq_length=max_seq_length,
                    load_in_4bit=False,
+                    trust_remote_code=trust_remote_code,
                )

            elif self.is_vision:
@ -226,6 +232,7 @@ class ExportBackend:
                    max_seq_length=max_seq_length,
                    dtype=None,
                    load_in_4bit=load_in_4bit,
+                    trust_remote_code=trust_remote_code,
                )
                tokenizer = processor  # For vision models, processor acts as tokenizer

@ -236,6 +243,7 @@ class ExportBackend:
                    max_seq_length=max_seq_length,
                    dtype=None,
                    load_in_4bit=load_in_4bit,
+                    trust_remote_code=trust_remote_code,
                )

            # Check if PEFT model
--- a/studio/backend/core/export/orchestrator.py
+++ b/studio/backend/core/export/orchestrator.py
@ -211,6 +211,7 @@ class ExportOrchestrator:
        checkpoint_path: str,
        max_seq_length: int = 2048,
        load_in_4bit: bool = True,
+        trust_remote_code: bool = False,
    ) -> Tuple[bool, str]:
        """Load a checkpoint for export.

@ -225,6 +226,7 @@ class ExportOrchestrator:
            "checkpoint_path": checkpoint_path,
            "max_seq_length": max_seq_length,
            "load_in_4bit": load_in_4bit,
+            "trust_remote_code": trust_remote_code,
        }

        # Always kill existing subprocess and spawn fresh.
--- a/studio/backend/core/export/worker.py
+++ b/studio/backend/core/export/worker.py
@ -84,6 +84,7 @@ def _handle_load(backend, cmd: dict, resp_queue: Any) -> None:
    checkpoint_path = cmd["checkpoint_path"]
    max_seq_length = cmd.get("max_seq_length", 2048)
    load_in_4bit = cmd.get("load_in_4bit", True)
+    trust_remote_code = cmd.get("trust_remote_code", False)

    try:
        _send_response(resp_queue, {
@ -96,6 +97,7 @@ def _handle_load(backend, cmd: dict, resp_queue: Any) -> None:
            checkpoint_path=checkpoint_path,
            max_seq_length=max_seq_length,
            load_in_4bit=load_in_4bit,
+            trust_remote_code=trust_remote_code,
        )

        _send_response(resp_queue, {
--- a/studio/backend/core/inference/inference.py
+++ b/studio/backend/core/inference/inference.py
@ -63,7 +63,8 @@ class InferenceBackend:
                   max_seq_length: int = 2048,
                   dtype = None,
                   load_in_4bit: bool = True,
-                   hf_token: Optional[str] = None) -> bool:
+                   hf_token: Optional[str] = None,
+                   trust_remote_code: bool = False) -> bool:
        """
        Load any model: base, LoRA adapter, text, or vision.
        """
@ -110,6 +111,7 @@ class InferenceBackend:
                        auto_model=CsmForConditionalGeneration,
                        load_in_4bit=False,
                        token=hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code=trust_remote_code,
                    )
                    FastModel.for_inference(model)
                    self.models[model_name]["model"] = model
@ -139,6 +141,7 @@ class InferenceBackend:
                            dtype=torch.float32,
                            load_in_4bit=False,
                            token=hf_token if hf_token and hf_token.strip() else None,
+                            trust_remote_code=trust_remote_code,
                        )
                    else:
                        # Base model: download full HF repo, then load from /LLM subfolder
@ -155,6 +158,7 @@ class InferenceBackend:
                            dtype=torch.float32,
                            load_in_4bit=False,
                            token=hf_token if hf_token and hf_token.strip() else None,
+                            trust_remote_code=trust_remote_code,
                        )

                    FastModel.for_inference(model)
@ -169,6 +173,7 @@ class InferenceBackend:
                        max_seq_length=max_seq_length,
                        load_in_4bit=False,
                        token=hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code=trust_remote_code,
                    )
                    FastModel.for_inference(model)
                    self.models[model_name]["model"] = model
@ -184,6 +189,7 @@ class InferenceBackend:
                        whisper_task="transcribe",
                        load_in_4bit=False,
                        token=hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code=trust_remote_code,
                    )
                    FastModel.for_inference(model)
                    model.eval()
@ -209,6 +215,7 @@ class InferenceBackend:
                        max_seq_length=max_seq_length,
                        load_in_4bit=False,
                        token=hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code=trust_remote_code,
                    )
                    FastLanguageModel.for_inference(model)
                    self.models[model_name]["model"] = model
@ -240,6 +247,7 @@ class InferenceBackend:
                    dtype=dtype,
                    load_in_4bit=load_in_4bit,
                    token=hf_token if hf_token and hf_token.strip() else None,
+                    trust_remote_code=trust_remote_code,
                )

                # Apply inference optimization
@ -270,6 +278,7 @@ class InferenceBackend:
                    processor = AutoProcessor.from_pretrained(
                        processor_source,
                        token=hf_token if hf_token and hf_token.strip() else None,
+                        trust_remote_code=trust_remote_code,
                    )
                    logger.info(f"Loaded {type(processor).__name__} from {processor_source}")

@ -285,6 +294,7 @@ class InferenceBackend:
                    dtype=dtype,
                    load_in_4bit=load_in_4bit,
                    token=hf_token if hf_token and hf_token.strip() else None,
+                    trust_remote_code=trust_remote_code,
                )

                # Apply inference optimization
--- a/studio/backend/core/inference/orchestrator.py
+++ b/studio/backend/core/inference/orchestrator.py
@ -258,6 +258,7 @@ class InferenceOrchestrator:
        dtype=None,
        load_in_4bit: bool = True,
        hf_token: Optional[str] = None,
+        trust_remote_code: bool = False,
    ) -> bool:
        """Load a model for inference.

@ -282,6 +283,7 @@ class InferenceOrchestrator:
                "load_in_4bit": load_in_4bit,
                "hf_token": hf_token or "",
                "gguf_variant": getattr(config, "gguf_variant", None),
+                "trust_remote_code": trust_remote_code,
            }

            # Always kill existing subprocess and spawn fresh.
--- a/studio/backend/core/inference/worker.py
+++ b/studio/backend/core/inference/worker.py
@ -155,6 +155,7 @@ def _handle_load(backend, config: dict, resp_queue: Any) -> None:
            max_seq_length=config.get("max_seq_length", 2048),
            load_in_4bit=load_in_4bit,
            hf_token=hf_token,
+            trust_remote_code=config.get("trust_remote_code", False),
        )

        if success:
--- a/studio/backend/core/training/trainer.py
+++ b/studio/backend/core/training/trainer.py
@ -329,9 +329,11 @@ class UnslothTrainer:
                   load_in_4bit: bool = True,
                   hf_token: Optional[str] = None,
                   is_dataset_image: bool = False,
-                   is_dataset_audio: bool = False) -> bool:
+                   is_dataset_audio: bool = False,
+                   trust_remote_code: bool = False) -> bool:
        """Load model for training (supports both text and vision models)"""
        self.load_in_4bit = load_in_4bit  # Store for training_meta.json
+        self.trust_remote_code = trust_remote_code  # For AutoProcessor etc. used during training
        try:
            if self.model is not None:
                del self.model
@ -450,6 +452,7 @@ class UnslothTrainer:
                    auto_model=CsmForConditionalGeneration,
                    load_in_4bit=False,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded CSM audio model")

@ -465,6 +468,7 @@ class UnslothTrainer:
                    whisper_language="English",
                    whisper_task="transcribe",
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                # Configure generation settings (notebook lines 100-105)
                self.model.generation_config.language = "<|en|>"
@ -481,6 +485,7 @@ class UnslothTrainer:
                    dtype=None,
                    load_in_4bit=load_in_4bit,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info(f"Loaded {self._audio_type} audio model (FastLanguageModel)")

@ -514,6 +519,7 @@ class UnslothTrainer:
                    dtype=torch.float32,  # Spark-TTS requires float32
                    load_in_4bit=False,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded Spark-TTS (bicodec) model")

@ -525,6 +531,7 @@ class UnslothTrainer:
                    max_seq_length=max_seq_length,
                    load_in_4bit=False,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded OuteTTS (dac) model (FastModel)")

@ -538,6 +545,7 @@ class UnslothTrainer:
                    dtype=None,
                    load_in_4bit=load_in_4bit,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded audio VLM model (FastModel)")

@ -549,6 +557,7 @@ class UnslothTrainer:
                    dtype=None,  # Auto-detect
                    load_in_4bit=load_in_4bit,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded vision model")

@ -568,6 +577,7 @@ class UnslothTrainer:
                    dtype=None,  # Auto-detect
                    load_in_4bit=load_in_4bit,
                    token=hf_token,
+                    trust_remote_code=trust_remote_code,
                )
                logger.info("Loaded text model")

@ -588,7 +598,7 @@ class UnslothTrainer:
                self._source_code_retried = True
                print(f"\n'could not get source code' — retrying once...\n")
                return self.load_model(model_name, max_seq_length, load_in_4bit, hf_token,
-                                       is_dataset_image, is_dataset_audio)
+                                       is_dataset_image, is_dataset_audio, trust_remote_code)
            error_msg = str(e)
            error_lower = error_msg.lower()
            if any(k in error_lower for k in ("gated repo", "access to it at", "401", "403", "unauthorized", "forbidden")):
@ -987,7 +997,10 @@ class UnslothTrainer:
        from datasets import Audio
        import torch

-        processor = AutoProcessor.from_pretrained(self.model_name)
+        processor = AutoProcessor.from_pretrained(
+            self.model_name,
+            trust_remote_code=getattr(self, "trust_remote_code", False),
+        )

        # Strip pad_to_multiple_of from tokenizer init_kwargs — fine-tuned models
        # (e.g. keanteng/sesame-csm-elise) save it in tokenizer_config.json, and
--- a/studio/backend/core/training/training.py
+++ b/studio/backend/core/training/training.py
@ -176,6 +176,7 @@ class TrainingBackend:
            "wandb_project": kwargs.get("wandb_project", "unsloth-training"),
            "enable_tensorboard": kwargs.get("enable_tensorboard", False),
            "tensorboard_dir": kwargs.get("tensorboard_dir", "runs"),
+            "trust_remote_code": kwargs.get("trust_remote_code", False),
        }

        # Derive load_in_4bit from training_type
--- a/studio/backend/core/training/worker.py
+++ b/studio/backend/core/training/worker.py
@ -192,14 +192,16 @@ def run_training_process(
            hf_token=hf_token,
            is_dataset_image=config.get("is_dataset_image", False),
            is_dataset_audio=config.get("is_dataset_audio", False),
+            trust_remote_code=config.get("trust_remote_code", False),
        )
        if not success or trainer.should_stop:
            if trainer.should_stop:
                event_queue.put({"type": "complete", "output_dir": None, "ts": time.time()})
            else:
+                error_msg = trainer.training_progress.error or "Failed to load model"
                event_queue.put({
                    "type": "error",
-                    "error": trainer.training_progress.error or "Failed to load model",
+                    "error": error_msg,
                    "stack": "", "ts": time.time(),
                })
            return
--- a/studio/backend/models/export.py
+++ b/studio/backend/models/export.py
@ -19,6 +19,10 @@ class LoadCheckpointRequest(BaseModel):
        True,
        description="Whether to load the model in 4-bit quantization",
    )
+    trust_remote_code: bool = Field(
+        False,
+        description="Allow loading models with custom code. Only enable for checkpoints/base models you trust.",
+    )


 class ExportStatusResponse(BaseModel):
--- a/studio/backend/models/inference.py
+++ b/studio/backend/models/inference.py
@ -18,6 +18,10 @@ class LoadRequest(BaseModel):
    load_in_4bit: bool = Field(True, description="Load model in 4-bit quantization")
    is_lora: bool = Field(False, description="Whether this is a LoRA adapter")
    gguf_variant: Optional[str] = Field(None, description="GGUF quantization variant (e.g. 'Q4_K_M')")
+    trust_remote_code: bool = Field(
+        False,
+        description="Allow loading models with custom code (e.g. NVIDIA Nemotron). Only enable for repos you trust.",
+    )


 class UnloadRequest(BaseModel):
--- a/studio/backend/models/training.py
+++ b/studio/backend/models/training.py
@ -13,6 +13,10 @@ class TrainingStartRequest(BaseModel):
    hf_token: Optional[str] = Field(None, description="HuggingFace token")
    load_in_4bit: bool = Field(True, description="Load model in 4-bit quantization")
    max_seq_length: int = Field(2048, description="Maximum sequence length")
+    trust_remote_code: bool = Field(
+        False,
+        description="Allow loading models with custom code (e.g. NVIDIA Nemotron). Only enable for repos you trust.",
+    )

    # Dataset parameters
    hf_dataset: Optional[str] = Field(None, description="HuggingFace dataset identifier")
--- a/studio/backend/routes/export.py
+++ b/studio/backend/routes/export.py
@ -102,6 +102,7 @@ async def load_checkpoint(
            checkpoint_path=request.checkpoint_path,
            max_seq_length=request.max_seq_length,
            load_in_4bit=request.load_in_4bit,
+            trust_remote_code=request.trust_remote_code,
        )

        if not success:
--- a/studio/backend/routes/inference.py
+++ b/studio/backend/routes/inference.py
@ -26,6 +26,7 @@ try:
    from core.inference.llama_cpp import LlamaCppBackend
    from utils.models import ModelConfig
    from utils.inference import load_inference_config
+    from utils.models.model_config import load_model_defaults
 except ImportError:
    parent_backend = backend_path.parent / "backend"
    if str(parent_backend) not in sys.path:
@ -34,6 +35,7 @@ except ImportError:
    from core.inference.llama_cpp import LlamaCppBackend
    from utils.models import ModelConfig
    from utils.inference import load_inference_config
+    from utils.models.model_config import load_model_defaults

 from models.inference import (
    LoadRequest,
@ -228,9 +230,22 @@ async def load_model(
            max_seq_length=request.max_seq_length,
            load_in_4bit=load_in_4bit,
            hf_token=request.hf_token,
+            trust_remote_code=request.trust_remote_code,
        )

        if not success:
+            # Check if YAML says this model needs trust_remote_code
+            if not request.trust_remote_code:
+                model_defaults = load_model_defaults(config.identifier)
+                yaml_trust = model_defaults.get("inference", {}).get("trust_remote_code", False)
+                if yaml_trust:
+                    raise HTTPException(
+                        status_code=400,
+                        detail=(
+                            f"Model '{config.display_name}' requires trust_remote_code to be enabled. "
+                            f"Please enable 'Trust remote code' in Chat Settings and try again."
+                        ),
+                    )
            raise HTTPException(
                status_code=500,
                detail=f"Failed to load model: {config.display_name}"
--- a/studio/backend/routes/training.py
+++ b/studio/backend/routes/training.py
@ -19,12 +19,14 @@ if str(backend_path) not in sys.path:
 # Import backend functions
 try:
    from core.training import get_training_backend
+    from utils.models.model_config import load_model_defaults
 except ImportError:
    # Fallback: try to import from parent directory
    parent_backend = backend_path.parent / "backend"
    if str(parent_backend) not in sys.path:
        sys.path.insert(0, str(parent_backend))
    from core.training import get_training_backend
+    from utils.models.model_config import load_model_defaults

 # Auth
 from auth.authentication import get_current_subject
@ -200,8 +202,19 @@ async def start_training(
            "wandb_project": request.wandb_project or "",
            "enable_tensorboard": request.enable_tensorboard,
            "tensorboard_dir": request.tensorboard_dir or "",
+            "trust_remote_code": request.trust_remote_code,
        }

+        # Training page has no trust_remote_code toggle — the value comes from
+        # YAML model defaults applied when the user selects a model.  As a safety
+        # net, consult the YAML directly so models that need it always get it.
+        if not training_kwargs["trust_remote_code"]:
+            model_defaults = load_model_defaults(request.model_name)
+            yaml_trust = model_defaults.get("training", {}).get("trust_remote_code", False)
+            if yaml_trust:
+                logger.info(f"YAML config sets trust_remote_code=True for {request.model_name}")
+                training_kwargs["trust_remote_code"] = True
+
        # Free GPU memory: shut down any running inference/export subprocesses
        # before training starts (they'd compete for VRAM otherwise)
        try:
--- a/studio/backend/utils/inference/inference_config.py
+++ b/studio/backend/utils/inference/inference_config.py
@ -59,6 +59,7 @@ def load_inference_config(model_identifier: str) -> Dict[str, Any]:
        "top_p": model_inference.get("top_p", default_inference.get("top_p", 0.95)),
        "top_k": model_inference.get("top_k", default_inference.get("top_k", -1)),
        "min_p": model_inference.get("min_p", default_inference.get("min_p", 0.01)),
+        "trust_remote_code": model_inference.get("trust_remote_code", default_inference.get("trust_remote_code", False)),
    }
    
    return inference_config
--- a/studio/frontend/src/config/training.ts
+++ b/studio/frontend/src/config/training.ts
@ -114,6 +114,7 @@ export const DEFAULT_HYPERPARAMS = {
  enableTensorboard: false,
  tensorboardDir: "runs",
  logFrequency: 10,
+  trustRemoteCode: false,
  finetuneVisionLayers: true,
  finetuneLanguageLayers: true,
  finetuneAttentionModules: true,
--- a/studio/frontend/src/features/chat/chat-settings-sheet.tsx
+++ b/studio/frontend/src/features/chat/chat-settings-sheet.tsx
@ -170,6 +170,7 @@ export function ChatSettingsPanel({
        ...p.params,
        systemPrompt: params.systemPrompt,
        checkpoint: params.checkpoint,
+        trustRemoteCode: params.trustRemoteCode,
      });
      setActivePreset(name);
    }
@ -330,17 +331,31 @@ export function ChatSettingsPanel({
          </CollapsibleSection>

          <CollapsibleSection icon={Settings02Icon} label="Settings">
-            <div className="flex items-center justify-between gap-3 py-1">
-              <div className="min-w-0">
-                <div className="text-xs font-medium">Auto title</div>
-                <div className="text-[11px] text-muted-foreground">
-                  Generate short title after reply.
+            <div className="flex flex-col gap-3 py-1">
+              <div className="flex items-center justify-between gap-3">
+                <div className="min-w-0">
+                  <div className="text-xs font-medium">Auto title</div>
+                  <div className="text-[11px] text-muted-foreground">
+                    Generate short title after reply.
+                  </div>
                </div>
+                <Switch
+                  checked={autoTitle}
+                  onCheckedChange={onAutoTitleChange}
+                />
+              </div>
+              <div className="flex items-center justify-between gap-3">
+                <div className="min-w-0">
+                  <div className="text-xs font-medium">Trust remote code</div>
+                  <div className="text-[11px] text-muted-foreground">
+                    Allow models with custom code (e.g. Nemotron). Only enable for repos you trust.
+                  </div>
+                </div>
+                <Switch
+                  checked={params.trustRemoteCode ?? false}
+                  onCheckedChange={set("trustRemoteCode")}
+                />
              </div>
-              <Switch
-                checked={autoTitle}
-                onCheckedChange={onAutoTitleChange}
-              />
            </div>
          </CollapsibleSection>
        </div>
--- a/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
+++ b/studio/frontend/src/features/chat/hooks/use-chat-model-runtime.ts
@ -124,6 +124,10 @@ function mergeRecommendedInference(
    topP: toFiniteNumber(inference?.top_p) ?? current.topP,
    topK: toFiniteNumber(inference?.top_k) ?? current.topK,
    minP: toFiniteNumber(inference?.min_p) ?? current.minP,
+    trustRemoteCode:
+      typeof inference?.trust_remote_code === "boolean"
+        ? inference.trust_remote_code
+        : current.trustRemoteCode,
  };
 }

@ -232,6 +236,7 @@ export function useChatModelRuntime() {
              previousWasUnloaded = true;
            }

+            const paramsBeforeLoad = useChatRuntimeStore.getState().params;
            const loadResponse = await loadModel({
              model_path: modelId,
              hf_token: null,
@ -239,6 +244,7 @@ export function useChatModelRuntime() {
              load_in_4bit: true,
              is_lora: isLora,
              gguf_variant: ggufVariant ?? null,
+              trust_remote_code: paramsBeforeLoad.trustRemoteCode ?? false,
            });

            const currentParams = useChatRuntimeStore.getState().params;
--- a/studio/frontend/src/features/chat/types/api.ts
+++ b/studio/frontend/src/features/chat/types/api.ts
@ -34,6 +34,8 @@ export interface LoadModelRequest {
  load_in_4bit: boolean;
  is_lora: boolean;
  gguf_variant?: string | null;
+  /** Allow loading models with custom code (e.g. NVIDIA Nemotron). Only enable for repos you trust. */
+  trust_remote_code?: boolean;
 }

 export interface ValidateModelResponse {
@ -74,6 +76,7 @@ export interface LoadModelResponse {
    top_p?: number;
    top_k?: number;
    min_p?: number;
+    trust_remote_code?: boolean;
  };
 }

--- a/studio/frontend/src/features/chat/types/runtime.ts
+++ b/studio/frontend/src/features/chat/types/runtime.ts
@ -7,6 +7,8 @@ export interface InferenceParams {
  maxTokens: number;
  systemPrompt: string;
  checkpoint: string;
+  /** Allow loading models with custom code (e.g. NVIDIA Nemotron). Only enable for repos you trust. */
+  trustRemoteCode?: boolean;
 }

 export const DEFAULT_INFERENCE_PARAMS: InferenceParams = {
@ -18,6 +20,7 @@ export const DEFAULT_INFERENCE_PARAMS: InferenceParams = {
  maxTokens: 4092,
  systemPrompt: "",
  checkpoint: "",
+  trustRemoteCode: false,
 };

 export interface ChatModelSummary {
--- a/studio/frontend/src/features/export/api/export-api.ts
+++ b/studio/frontend/src/features/export/api/export-api.ts
@ -50,6 +50,8 @@ export async function loadCheckpoint(params: {
  checkpoint_path: string;
  max_seq_length?: number;
  load_in_4bit?: boolean;
+  /** Allow loading models with custom code. Only enable for checkpoints you trust. */
+  trust_remote_code?: boolean;
 }): Promise<ExportOperationResponse> {
  const response = await authFetch("/api/export/load-checkpoint", {
    method: "POST",
--- a/studio/frontend/src/features/training/api/mappers.ts
+++ b/studio/frontend/src/features/training/api/mappers.ts
@ -36,6 +36,7 @@ export function buildTrainingStartPayload(
    hf_token: config.hfToken.trim() || null,
    load_in_4bit: adapterMethod ? isQloraMethod : false,
    max_seq_length: config.contextLength,
+    trust_remote_code: config.trustRemoteCode ?? false,
    hf_dataset: hfDataset,
    subset: hfDataset ? config.datasetSubset : null,
    train_split: hfDataset ? config.datasetSplit : null,
--- a/studio/frontend/src/features/training/api/models-api.ts
+++ b/studio/frontend/src/features/training/api/models-api.ts
@ -22,6 +22,7 @@ interface BackendTrainingDefaults {
  packing?: boolean;
  train_on_completions?: boolean;
  gradient_checkpointing?: "none" | "true" | "unsloth";
+  trust_remote_code?: boolean;
 }

 interface BackendLoraDefaults {
--- a/studio/frontend/src/features/training/lib/model-defaults.ts
+++ b/studio/frontend/src/features/training/lib/model-defaults.ts
@ -30,6 +30,7 @@ type ModelDefaultsPatch = Partial<
    | "tensorboardDir"
    | "logFrequency"
    | "finetuneVisionLayers"
+    | "trustRemoteCode"
    | "finetuneLanguageLayers"
    | "finetuneAttentionModules"
    | "finetuneMLPModules"
@ -133,6 +134,9 @@ export function mapBackendModelConfigToTrainingPatch(
    patch.gradientCheckpointing = gradientCheckpointing;
  }

+  const trustRemoteCode = toBoolean(training?.trust_remote_code);
+  if (trustRemoteCode !== undefined) patch.trustRemoteCode = trustRemoteCode;
+
  const loraRank = toNumber(lora?.lora_r);
  if (loraRank !== undefined) patch.loraRank = loraRank;

--- a/studio/frontend/src/features/training/types/api.ts
+++ b/studio/frontend/src/features/training/types/api.ts
@ -4,6 +4,8 @@ export interface TrainingStartRequest {
  hf_token: string | null;
  load_in_4bit: boolean;
  max_seq_length: number;
+  /** Allow loading models with custom code. Only enable for repos you trust. */
+  trust_remote_code?: boolean;
  hf_dataset: string | null;
  subset: string | null;
  train_split: string | null;
--- a/studio/frontend/src/features/training/types/config.ts
+++ b/studio/frontend/src/features/training/types/config.ts
@ -63,6 +63,7 @@ export interface TrainingConfigState {
  isCheckingDataset: boolean;
  isDatasetImage: boolean | null;
  isDatasetAudio: boolean;
+  trustRemoteCode: boolean;
  finetuneVisionLayers: boolean;
  finetuneLanguageLayers: boolean;
  finetuneAttentionModules: boolean;