See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigscience/bloomz-560m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 73614267d6b5037b_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/73614267d6b5037b_train_data.json
  type:
    field_instruction: instruction
    field_output: response
    format: '{instruction}'
    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: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/0967fbc7-e04c-42b3-89c8-59d87c24ac11
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: null
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 6528
micro_batch_size: 4
mlflow_experiment_name: /tmp/73614267d6b5037b_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: 1024
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: ebd855c9-d988-4fdb-b54f-62851b251b61
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ebd855c9-d988-4fdb-b54f-62851b251b61
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

0967fbc7-e04c-42b3-89c8-59d87c24ac11

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: 2.3074

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: 3722

Training results

Training Loss	Epoch	Step	Validation Loss
24.9657	0.0005	1	3.0363
19.2565	0.0537	100	2.5059
18.9865	0.1075	200	2.4698
22.3663	0.1612	300	2.4533
15.2299	0.2150	400	2.4415
22.0234	0.2687	500	2.4282
19.6186	0.3225	600	2.4188
18.2559	0.3762	700	2.4128
18.005	0.4300	800	2.4044
20.5376	0.4837	900	2.3974
17.9586	0.5375	1000	2.3906
19.3092	0.5912	1100	2.3820
20.6819	0.6449	1200	2.3766
18.083	0.6987	1300	2.3701
19.7011	0.7524	1400	2.3603
17.9218	0.8062	1500	2.3576
19.294	0.8599	1600	2.3524
19.6344	0.9137	1700	2.3478
17.7207	0.9674	1800	2.3432
18.2821	1.0214	1900	2.3417
17.7064	1.0752	2000	2.3362
17.5435	1.1289	2100	2.3326
18.7645	1.1827	2200	2.3291
17.0431	1.2364	2300	2.3272
17.5376	1.2902	2400	2.3240
18.8583	1.3439	2500	2.3194
17.6271	1.3976	2600	2.3191
18.444	1.4514	2700	2.3139
16.9473	1.5051	2800	2.3115
16.8502	1.5589	2900	2.3099
16.3382	1.6126	3000	2.3090
17.5823	1.6664	3100	2.3066
17.821	1.7201	3200	2.3072
18.5151	1.7739	3300	2.3062
17.5566	1.8276	3400	2.3066
19.8288	1.8814	3500	2.3074

Framework versions

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