See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: echarlaix/tiny-random-mistral
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 1d4c9f6e3fdcd125_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/1d4c9f6e3fdcd125_train_data.json
  type:
    field_instruction: instruction
    field_output: output
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
early_stopping_threshold: 0.0001
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: romainnn/f8960a9d-edeb-4b29-bcdb-f5db588483bc
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_grad_norm: 1.0
max_steps: 2652
micro_batch_size: 4
mlflow_experiment_name: /tmp/1d4c9f6e3fdcd125_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
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 26545dad-1c8b-4b71-bd6a-45e32355614f
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 26545dad-1c8b-4b71-bd6a-45e32355614f
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

f8960a9d-edeb-4b29-bcdb-f5db588483bc

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

• Loss: 10.2714

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: 2652

Training results

Training Loss	Epoch	Step	Validation Loss
83.0589	0.0007	1	10.3802
82.5901	0.0686	100	10.3337
82.5677	0.1371	200	10.3225
82.521	0.2057	300	10.3162
82.3826	0.2742	400	10.3090
82.4453	0.3428	500	10.3018
82.4075	0.4113	600	10.2958
82.3576	0.4799	700	10.2901
82.3485	0.5485	800	10.2865
82.3176	0.6170	900	10.2842
82.255	0.6856	1000	10.2815
82.195	0.7541	1100	10.2797
82.2751	0.8227	1200	10.2779
82.3208	0.8913	1300	10.2764
82.2335	0.9598	1400	10.2753
82.2106	1.0287	1500	10.2744
82.2033	1.0973	1600	10.2736
82.223	1.1658	1700	10.2730
82.3021	1.2344	1800	10.2725
82.313	1.3029	1900	10.2722
82.3386	1.3715	2000	10.2719
82.2007	1.4401	2100	10.2718
82.2385	1.5086	2200	10.2716
82.2108	1.5772	2300	10.2715
82.254	1.6457	2400	10.2714
82.2093	1.7143	2500	10.2714
82.2733	1.7828	2600	10.2714

Framework versions

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