See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-1B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - afce925b7308b26a_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/afce925b7308b26a_train_data.json
  type:
    field_input: dialogue
    field_instruction: template_name
    field_output: summary
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 50
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: true
group_by_length: false
hub_model_id: romainnn/a0d17194-207b-4a6e-b479-89a27e2c6965
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_steps: 2073
micro_batch_size: 4
mlflow_experiment_name: /tmp/afce925b7308b26a_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 2048
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 94e81d74-a532-486d-86fb-6485bdfbcddb
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 94e81d74-a532-486d-86fb-6485bdfbcddb
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

a0d17194-207b-4a6e-b479-89a27e2c6965

This model is a fine-tuned version of unsloth/Llama-3.2-1B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1983

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• 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
• training_steps: 2073

Training results

Training Loss	Epoch	Step	Validation Loss
2.4706	0.0006	1	2.5195
1.2327	0.0294	50	1.2823
1.1913	0.0588	100	1.2301
1.2492	0.0883	150	1.1923
1.155	0.1177	200	1.1574
1.1591	0.1471	250	1.1221
1.1899	0.1765	300	1.0899
1.0132	0.2060	350	1.0607
0.9813	0.2354	400	1.0220
1.0035	0.2648	450	0.9830
1.109	0.2942	500	0.9482
1.0469	0.3236	550	0.9101
0.7811	0.3531	600	0.8685
0.8326	0.3825	650	0.8316
0.7991	0.4119	700	0.7946
0.755	0.4413	750	0.7587
0.7753	0.4708	800	0.7208
0.7614	0.5002	850	0.6819
0.552	0.5296	900	0.6446
0.6636	0.5590	950	0.6092
0.5829	0.5885	1000	0.5794
0.4899	0.6179	1050	0.5421
0.5464	0.6473	1100	0.5072
0.4774	0.6767	1150	0.4716
0.4009	0.7061	1200	0.4413
0.4265	0.7356	1250	0.4094
0.3909	0.7650	1300	0.3790
0.3691	0.7944	1350	0.3555
0.4325	0.8238	1400	0.3314
0.3545	0.8533	1450	0.3099
0.2653	0.8827	1500	0.2868
0.2634	0.9121	1550	0.2700
0.3145	0.9415	1600	0.2529
0.2409	0.9709	1650	0.2376
0.3415	1.0004	1700	0.2250
0.1703	1.0298	1750	0.2167
0.1621	1.0592	1800	0.2107
0.1723	1.0886	1850	0.2049
0.1343	1.1181	1900	0.2018
0.1684	1.1475	1950	0.1997
0.1737	1.1769	2000	0.1988
0.1514	1.2063	2050	0.1983

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1