See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: echarlaix/tiny-random-PhiForCausalLM
bf16: auto
chat_template: phi_3
dataset_prepared_path: null
datasets:
- data_files:
  - 1c3359627c73674a_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/1c3359627c73674a_train_data.json
  type:
    field_input: about_book
    field_instruction: topic_name
    field_output: conversation
    format: '{instruction} {input}'
    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: 1
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: false
group_by_length: false
hub_model_id: error577/86c41dc0-e58a-448b-9dd3-f03357b788a0
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.001
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
#max_steps: 100
micro_batch_size: 4
mlflow_experiment_name: /tmp/1c3359627c73674a_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 24
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
saves_per_epoch: 1
sequence_len: 4096
special_tokens:
  pad_token: <|endoftext|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: ba48bfb0-9311-44bf-bd5d-53c685694e8d
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ba48bfb0-9311-44bf-bd5d-53c685694e8d
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

86c41dc0-e58a-448b-9dd3-f03357b788a0

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

• Loss: 6.7952

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: 0.001
• 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
• num_epochs: 24

Training results

Training Loss	Epoch	Step	Validation Loss
6.9369	0.0096	1	6.9385
6.8626	1.0	104	6.8637
6.8467	2.0	208	6.8443
6.8355	3.0	312	6.8299
6.8247	4.0	416	6.8210
6.8202	5.0	520	6.8154
6.8188	6.0	624	6.8110
6.8109	7.0	728	6.8078
6.821	8.0	832	6.8050
6.8093	9.0	936	6.8028
6.8046	10.0	1040	6.8014
6.8132	11.0	1144	6.8002
6.8058	12.0	1248	6.7990
6.8112	13.0	1352	6.7982
6.8054	14.0	1456	6.7974
6.8078	15.0	1560	6.7969
6.8045	16.0	1664	6.7963
6.8032	17.0	1768	6.7961
6.8015	18.0	1872	6.7957
6.8004	19.0	1976	6.7955
6.8054	20.0	2080	6.7953
6.8052	21.0	2184	6.7952
6.8041	22.0	2288	6.7952
6.8051	23.0	2392	6.7952
6.8056	24.0	2496	6.7952

Framework versions

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