XavierThibaudon/anomaly-analyzer-lora

See axolotl config

axolotl version: 0.11.0.dev0

# Axolotl LoRA Fine-tuning Config for Anomaly Analyzer
# 
# Train a specialized model for crypto exchange anomaly analysis.
# Based on Llama 3.2 1B - small enough for fast inference, trainable on consumer GPUs.

base_model: meta-llama/Llama-3.2-1B-Instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

# Llama 3.2 lacks a pad token — use the built-in finetune pad token
special_tokens:
  pad_token: "<|finetune_right_pad_id|>"

# Use 8-bit quantization for memory efficiency
load_in_8bit: true

# LoRA Configuration
adapter: lora
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj
  - k_proj
  - o_proj
  - gate_proj
  - up_proj
  - down_proj

# Dataset
datasets:
  #- path: /data/data/training-data-axolotl.jsonl
  - path: /data/data/training-data-combined.jsonl
    type: alpaca

# Output
output_dir: /data/lora-anomaly-analyzer

# Training Configuration
micro_batch_size: 2
gradient_accumulation_steps: 4
num_epochs: 3
learning_rate: 2e-4
lr_scheduler: cosine
warmup_ratio: 0.1
weight_decay: 0.01

# Optimizer
optimizer: adamw_torch
adam_beta1: 0.9
adam_beta2: 0.999

# Validation (disabled - too few examples for meaningful eval)
val_set_size: 0

# Logging
logging_steps: 10
save_steps: 100

# Mixed precision
bf16: auto
tf32: false

# Sequence length (our prompts can be long)
sequence_len: 2048
sample_packing: false

# Gradient checkpointing for memory
gradient_checkpointing: true

# Reproducibility
seed: 42

data/lora-anomaly-analyzer

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct on the /data/data/training-data-combined.jsonl 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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 8
• training_steps: 84

Training results

Framework versions

• PEFT 0.15.2
• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1