See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: HuggingFaceH4/tiny-random-LlamaForCausalLM
bf16: true
chat_template: llama3
cosine_min_lr_ratio: 0.3
dataset_prepared_path: null
datasets:
- data_files:
  - 94bc225e90b38509_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/94bc225e90b38509_train_data.json
  type:
    field_input: description
    field_instruction: reference
    field_output: article
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 4
eval_max_new_tokens: 128
eval_steps: 200
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/6dcbd096-39ee-4abe-93fb-d769dbdb5e0a
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: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 75317
micro_batch_size: 4
mlflow_experiment_name: /tmp/94bc225e90b38509_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 200
sequence_len: 2048
special_tokens:
  pad_token: </s>
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: b5890fb6-bacc-436a-9f1f-7bcb9df35f08
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: b5890fb6-bacc-436a-9f1f-7bcb9df35f08
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

6dcbd096-39ee-4abe-93fb-d769dbdb5e0a

This model is a fine-tuned version of HuggingFaceH4/tiny-random-LlamaForCausalLM on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.3030

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: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 3889

Training results

Training Loss	Epoch	Step	Validation Loss
10.375	0.0008	1	10.3741
10.33	0.1543	200	10.3258
10.3187	0.3085	400	10.3177
10.3149	0.4628	600	10.3138
10.3168	0.6170	800	10.3112
10.3168	0.7713	1000	10.3095
10.3172	0.9256	1200	10.3083
14.2661	1.0800	1400	10.3073
8.8355	1.2343	1600	10.3065
12.1893	1.3885	1800	10.3059
8.483	1.5428	2000	10.3053
10.4896	1.6971	2200	10.3048
11.2808	1.8513	2400	10.3044
6.5668	2.0058	2600	10.3044
10.2144	2.1600	2800	10.3040
11.3883	2.3143	3000	10.3036
11.115	2.4686	3200	10.3035
10.6235	2.6228	3400	10.3032
13.6615	2.7771	3600	10.3031
11.4675	2.9314	3800	10.3030

Framework versions

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