See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: JackFram/llama-160m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 805f058c8bba5205_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/805f058c8bba5205_train_data.json
  type:
    field_input: Title
    field_instruction: meshMajor
    field_output: abstractText
    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: 100
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/eccd3ba8-04ba-4ed0-a969-37f1272694ce
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: 128
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3060
micro_batch_size: 4
mlflow_experiment_name: /tmp/805f058c8bba5205_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
special_tokens:
  pad_token: </s>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
use_rslora: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 1e98fb5c-7909-450b-aaa0-df4dfe79858d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 1e98fb5c-7909-450b-aaa0-df4dfe79858d
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

eccd3ba8-04ba-4ed0-a969-37f1272694ce

This model is a fine-tuned version of JackFram/llama-160m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.1968

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: 8
• total_train_batch_size: 32
• 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: 2988

Training results

Training Loss	Epoch	Step	Validation Loss
2.4853	0.0007	1	2.4221
2.2173	0.0669	100	2.2804
2.1409	0.1339	200	2.2656
2.2481	0.2008	300	2.2580
2.2776	0.2677	400	2.2506
2.2549	0.3347	500	2.2453
2.1056	0.4016	600	2.2413
2.3304	0.4685	700	2.2368
2.2064	0.5355	800	2.2321
2.1778	0.6024	900	2.2298
2.1069	0.6693	1000	2.2251
2.1674	0.7363	1100	2.2210
2.2225	0.8032	1200	2.2184
2.1332	0.8701	1300	2.2165
2.2397	0.9371	1400	2.2129
2.1167	1.0040	1500	2.2104
2.1367	1.0710	1600	2.2096
2.1625	1.1379	1700	2.2077
2.3569	1.2048	1800	2.2061
2.0947	1.2718	1900	2.2044
2.0811	1.3387	2000	2.2027
2.1792	1.4056	2100	2.2018
2.1443	1.4726	2200	2.2001
2.0647	1.5395	2300	2.1993
2.1534	1.6064	2400	2.1983
2.1485	1.6734	2500	2.1978
2.1135	1.7403	2600	2.1973
2.1658	1.8072	2700	2.1971
2.1297	1.8742	2800	2.1967
2.0333	1.9411	2900	2.1968

Framework versions

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