See axolotl config

axolotl version: 0.4.1

adapter: qlora
base_model: llamafactory/tiny-random-Llama-3
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 8de1c9f60234da41_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/8de1c9f60234da41_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: output
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 1
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/e549a581-1920-4fcf-a8c9-a57c73863d30
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.0001
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_steps: 1000
micro_batch_size: 1
mlflow_experiment_name: /tmp/8de1c9f60234da41_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 50
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: true
saves_per_epoch: 1
sequence_len: 4096
special_tokens:
  pad_token: <|eot_id|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.02
wandb_entity: null
wandb_mode: online
wandb_name: 6ddc676b-9cb1-4ead-95d2-6c83ea27f192
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 6ddc676b-9cb1-4ead-95d2-6c83ea27f192
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

e549a581-1920-4fcf-a8c9-a57c73863d30

This model is a fine-tuned version of llamafactory/tiny-random-Llama-3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 11.7343

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• 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
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss
11.7679	0.0043	1	11.7672
11.7659	0.0856	20	11.7667
11.7657	0.1712	40	11.7659
11.7648	0.2568	60	11.7649
11.7632	0.3424	80	11.7634
11.7602	0.4280	100	11.7608
11.7552	0.5136	120	11.7564
11.7498	0.5993	140	11.7501
11.7441	0.6849	160	11.7450
11.7427	0.7705	180	11.7424
11.7396	0.8561	200	11.7412
11.7399	0.9417	220	11.7405
11.5384	1.0284	240	11.7399
12.5499	1.1140	260	11.7395
12.028	1.1996	280	11.7393
11.8707	1.2852	300	11.7390
11.9146	1.3708	320	11.7388
10.9729	1.4564	340	11.7385
11.9595	1.5420	360	11.7383
11.2912	1.6276	380	11.7380
12.0471	1.7132	400	11.7377
11.1845	1.7988	420	11.7373
11.999	1.8844	440	11.7369
11.4751	1.9700	460	11.7365
11.3704	2.0567	480	11.7362
11.2276	2.1423	500	11.7359
11.2259	2.2279	520	11.7358
11.2708	2.3135	540	11.7356
11.2495	2.3991	560	11.7355
11.947	2.4848	580	11.7353
11.885	2.5704	600	11.7352
11.9815	2.6560	620	11.7351
12.7368	2.7416	640	11.7350
12.1553	2.8272	660	11.7349
11.6847	2.9128	680	11.7348
21.76	3.0005	700	11.7347
11.4174	3.0861	720	11.7346
11.6541	3.1717	740	11.7346
11.6661	3.2574	760	11.7345
12.0102	3.3430	780	11.7345
11.8984	3.4286	800	11.7344
11.6288	3.5142	820	11.7344
12.14	3.5998	840	11.7343
12.1483	3.6854	860	11.7343
11.813	3.7710	880	11.7343
11.5873	3.8566	900	11.7343
11.8628	3.9422	920	11.7343
12.1773	4.0289	940	11.7343
12.5054	4.1145	960	11.7343
11.6244	4.2001	980	11.7343
11.5367	4.2857	1000	11.7343

Framework versions

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