shalem-2024/Qwen2.5-72B-dSPACE-XML-Adapter

See axolotl config

axolotl version: 0.4.1

# config.yaml - H200 STABLE EDITION (Root Folder)
base_model: Qwen/Qwen2.5-72B-Instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_4bit: true
strict: false
adapter: qlora
lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true

# DATASET (Stable Syntax)
datasets:
  - path: json
    data_files: shuffled_batch_1.jsonl
    type: alpaca

# MEMORY (H200 + Flash Attn)
sequence_len: 24576
sample_packing: true
pad_to_sequence_len: true

# TRAINING
gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0001

# OPTIMIZATION
train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: false
gradient_checkpointing: true
local_rank:
logging_steps: 1

# FLASH ATTENTION (Enabled)
flash_attention: true
xformers_attention: false

warmup_steps: 10
evals_per_epoch: 0
saves_per_epoch: 1
weight_decay: 0.0
output_dir: ./model-out

model-out

This model is a fine-tuned version of Qwen/Qwen2.5-72B-Instruct on the None dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Use adamw_bnb_8bit with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1