minpeter/QLoRA-Llama-3.2-1B-chatml-tool-v1

See axolotl config

axolotl version: 0.6.0

base_model: minpeter/Llama-3.2-1B-AlternateTokenizer-chatml

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
  - path: func-calling-singleturn.jsonl
    type: chat_template
    chat_template: chatml
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    shards: 2

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: 4096
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
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 is a fine-tuned version of minpeter/Llama-3.2-1B-AlternateTokenizer-chatml on the teknium/OpenHermes-2.5 and the func-calling-singleturn.jsonl datasets. It achieves the following results on the evaluation set:

• Loss: 0.9507

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
3.1156	0.0129	1	2.5490
1.2341	0.2589	20	1.2880
1.202	0.5178	40	1.0696
1.1038	0.7767	60	1.0076
1.0608	1.0259	80	0.9808
1.3812	1.2848	100	0.9619
0.9097	1.5437	120	0.9541
0.9775	1.8026	140	0.9507

Framework versions

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