added configs from Ollama

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

@@ -1,6 +1,7 @@
 # Model defaults for unsloth/codegemma-7b-bnb-4bit
 # Based on CodeGemma_(7B)-Conversational.ipynb
 # Also applies to: unsloth/codegemma-7b, google/codegemma-7b
+# added inference parameters from Ollama
 
 training:
   max_seq_length: 4096
@@ -44,3 +45,7 @@ logging:
   enable_tensorboard: false
   tensorboard_dir: "runs"
   log_frequency: 10
+
+inference:
+  temperature: 0
+  top_p: 0.9

studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml

@@ -1,6 +1,7 @@
 # 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
+# added inference parameters from Ollama
 
 training:
   max_seq_length: 1024
@@ -45,3 +46,8 @@ logging:
   tensorboard_dir: "runs"
   log_frequency: 10
 
+inference:
+  temperature: 0.6
+  top_k: 20
+  top_p: 0.95
+

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

@@ -1,6 +1,7 @@
 # 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
+# added inference parameters from Ollama
 
 training:
   max_seq_length: 2048
@@ -45,3 +46,8 @@ logging:
   tensorboard_dir: "runs"
   log_frequency: 10
 
+inference:
+  temperature: 0.6
+  top_k: 20
+  top_p: 0.95
+

studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml

@@ -1,6 +1,7 @@
 # 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
+# added inference parameters from Ollama
 
 training:
   max_seq_length: 2048
@@ -45,3 +46,8 @@ logging:
   tensorboard_dir: "runs"
   log_frequency: 10
 
+inference:
+  temperature: 0.6
+  top_k: 20
+  top_p: 0.95
+

studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml

@@ -1,6 +1,7 @@
 # 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
+# added inference parameters from Ollama
 
 training:
   max_seq_length: 2048
@@ -45,3 +46,8 @@ logging:
   tensorboard_dir: "runs"
   log_frequency: 10
 
+inference:
+  temperature: 0.6
+  top_k: 20
+  top_p: 0.95
+