mgfrantz/cmte-demo

See axolotl config

axolotl version: 0.4.1

base_model: meta-llama/Llama-3.2-3B

load_in_8bit: false
load_in_4bit: true
strict: false
adapter: qlora

# Data config
dataset_prepared_path: data
chat_template: llama3
datasets:
  - path: data/train.jsonl
    ds_type: json
    data_files:
      - data/train.jsonl
    type: sharegpt

test_datasets:
  - path: data/eval.jsonl
    ds_type: json
    # You need to specify a split. For "json" datasets the default split is called "train".
    split: train
    type: sharegpt
    data_files:
      - data/eval.jsonl

sequence_len: 4096
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 2
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|finetune_right_pad_id|>
  eos_token: <|eot_id|>

model-out

This model is a fine-tuned version of meta-llama/Llama-3.2-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5353

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: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.9615	0.0696	1	0.9211
0.8231	0.2783	4	0.9140
0.779	0.5565	8	0.8034
0.545	0.8348	12	0.6488
0.3712	1.1043	16	0.5825
0.336	1.3826	20	0.5492
0.3231	1.6609	24	0.5379
0.2729	1.9391	28	0.5353

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.4.0+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1