owusu/DS-LLM-TEMPLATE-FINETUNING
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

instruct model setup

Browse Source
This commit is contained in:
Your Name
2025-08-28 17:57:59 +00:00
parent 77c563f358
commit d49b4ff2d5
55 changed files with 27760 additions and 326 deletions
Download Patch File Download Diff File Expand all files Collapse all files
configs/instruct/.ipynb_checkpoints/sample-checkpoint.yaml
+3 -2
View File
@@ -34,7 +34,7 @@ data:
# Model Configuration
model:
name: "unsloth/Qwen2.5-Coder-7B" # Model name from HuggingFace Hub (optimized for instruction following)
name: "unsloth/Qwen2.5-14B-Instruct" # Model name from HuggingFace Hub (optimized for instruction following)
max_length: 2048 # Maximum sequence length for tokenization
max_seq_length: 2048 # Maximum sequence length for training (RoPE scaling supported)
dtype: null # Data type: null for auto detection, float16 for Tesla T4/V100, bfloat16 for Ampere+
@@ -42,7 +42,7 @@ model:
token: null # HuggingFace token for gated models (e.g., "hf_...")
# Training Model Parameters
training_model: "unsloth/Qwen2.5-Coder-7B" # Model to use for training
training_model: "unsloth/Qwen2.5-14B-Instruct" # Model to use for training
training_max_seq_length: 2048 # Max sequence length for training
training_dtype: null # Data type for training
training_load_in_4bit: true # 4bit quantization for training
@@ -67,6 +67,7 @@ training:
# Output Configuration
output_dir: "./outputs" # Directory for training checkpoints
save_name: "qwen_2.5_test"
model_output_dir: "./models/instruct" # Directory to save the trained model
# Inference Configuration
configs/instruct/sample.yaml
+3 -2
View File
@@ -34,7 +34,7 @@ data:
# Model Configuration
model:
name: "unsloth/Qwen2.5-Coder-7B" # Model name from HuggingFace Hub (optimized for instruction following)
name: "unsloth/Qwen2.5-14B-Instruct" # Model name from HuggingFace Hub (optimized for instruction following)
max_length: 2048 # Maximum sequence length for tokenization
max_seq_length: 2048 # Maximum sequence length for training (RoPE scaling supported)
dtype: null # Data type: null for auto detection, float16 for Tesla T4/V100, bfloat16 for Ampere+
@@ -42,7 +42,7 @@ model:
token: null # HuggingFace token for gated models (e.g., "hf_...")
# Training Model Parameters
training_model: "unsloth/Qwen2.5-Coder-7B" # Model to use for training
training_model: "unsloth/Qwen2.5-14B-Instruct" # Model to use for training
training_max_seq_length: 2048 # Max sequence length for training
training_dtype: null # Data type for training
training_load_in_4bit: true # 4bit quantization for training
@@ -67,6 +67,7 @@ training:
# Output Configuration
output_dir: "./outputs" # Directory for training checkpoints
save_name: "qwen_2.5_test"
model_output_dir: "./models/instruct" # Directory to save the trained model
# Inference Configuration