See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigscience/bloomz-560m
bf16: true
chat_template: llama3
datasets:
- data_files:
  - bddc5a91032831aa_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/bddc5a91032831aa_train_data.json
  type:
    field_input: ''
    field_instruction: raw_text
    field_output: clean_text
    format: '{instruction}'
    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: 4
flash_attention: false
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: true
group_by_length: false
hub_model_id: lesso04/b266b8dd-22fe-4d9d-9c81-d1e2e255f0c3
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0001
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
lr_scheduler: cosine
max_memory:
  0: 77GiB
max_steps: 50
micro_batch_size: 8
mlflow_experiment_name: /tmp/bddc5a91032831aa_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: 25
save_strategy: steps
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: true
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: a32a4946-251d-4e5b-8c9a-f58e97c6800d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a32a4946-251d-4e5b-8c9a-f58e97c6800d
warmup_steps: 10
weight_decay: 0.01
xformers_attention: false

b266b8dd-22fe-4d9d-9c81-d1e2e255f0c3

This model is a fine-tuned version of bigscience/bloomz-560m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5448

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
3.5773	0.0004	1	1.7676
3.5935	0.0022	5	1.7383
3.3434	0.0045	10	1.6409
3.2407	0.0067	15	1.6080
3.163	0.0089	20	1.5857
3.3937	0.0111	25	1.5708
2.9551	0.0134	30	1.5597
2.9878	0.0156	35	1.5517
3.2742	0.0178	40	1.5471
3.5747	0.0201	45	1.5454
3.1134	0.0223	50	1.5448

Framework versions

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