minpeter/QLoRA-Llama-3.2-1B-chatml-v3

See axolotl config

axolotl version: 0.6.0

base_model: "./llama-1b"
#base_model: teknium/Llama-3.1-AlternateTokenizer
#tokenizer_config: teknium/Llama-3.1-AlternateTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: teknium/OpenHermes-2.5
    type: chat_template
    chat_template: chatml
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    shards: 800

save_safetensors: true
auto_resume_from_checkpoints: false
save_steps: 200

chat_template: chatml
dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./output

adapter: qlora
lora_model_dir:

sequence_len: 2048
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

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

wandb_project: "axolotl"
wandb_entity: "kasfiekfs-e"
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
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:
  bos_token: <|begin_of_text|>
  eos_token: <|im_end|>
  pad_token: <|end_of_text|>

# <--- unsloth config --->
unsloth_lora_mlp: true
unsloth_lora_qkv: true
unsloth_lora_o: true

output

This model was trained from scratch on the teknium/OpenHermes-2.5 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9959

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_BNB 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: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.0246	0.0370	1	1.1798
1.2085	0.2593	7	1.1312
1.1683	0.5185	14	1.0579
1.0703	0.7778	21	1.0243
1.0815	1.0370	28	1.0119
1.2227	1.2963	35	1.0031
1.1577	1.5556	42	0.9977
1.1897	1.8148	49	0.9959

Framework versions

• PEFT 0.14.0
• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0