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

instruct mode added to pipiline

Browse Source
This commit is contained in:
Your Name
2025-08-28 16:46:24 +00:00
parent 78d519efbf
commit 77c563f358
16 changed files with 19404 additions and 161 deletions
Download Patch File Download Diff File Expand all files Collapse all files
configs/instruct/sample.yaml
+3 -3
View File
@@ -10,7 +10,7 @@ task:
# Data Processing Configuration
data:
source: "custom" # Data source: "huggingface" or "custom"
data_path: "./data/raw/instruct/code_reasoning.jsonl" # Path to conversation data file
data_path: "data/raw/swe_reasoning_dataset (3).jsonl" # Path to conversation data file
data_format: "jsonl" # Data format: "jsonl", "json"
# Field Mapping for Conversation Data
@@ -34,7 +34,7 @@ data:
# Model Configuration
model:
name: "unsloth/Qwen2.5-72B-Instruct" # Model name from HuggingFace Hub (optimized for instruction following)
name: "unsloth/Qwen2.5-Coder-7B" # 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/Qwen2.5-72B-Instruct" # Model to use for training
training_model: "unsloth/Qwen2.5-Coder-7B" # 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