See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: microsoft/phi-1_5
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - a3d7b5189cf022d9_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a3d7b5189cf022d9_train_data.json
  type:
    field_input: intent
    field_instruction: instruction
    field_output: response
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: false
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/85902018-8a10-4166-9ac5-ac7637c7c4c4
hub_repo: null
hub_strategy: null
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
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 3600
micro_batch_size: 4
mlflow_experiment_name: /tmp/a3d7b5189cf022d9_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
special_tokens:
  pad_token: <|endoftext|>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.03351206434316354
wandb_entity: null
wandb_mode: online
wandb_name: e9df8e98-2bbe-4eb3-b851-e60f3c690884
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: e9df8e98-2bbe-4eb3-b851-e60f3c690884
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

85902018-8a10-4166-9ac5-ac7637c7c4c4

This model is a fine-tuned version of microsoft/phi-1_5 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4561

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: 3600

Training results

Training Loss	Epoch	Step	Validation Loss
0.7478	0.0002	1	0.7768
0.5511	0.0222	100	0.5635
0.5676	0.0444	200	0.5444
0.5073	0.0666	300	0.5319
0.5047	0.0888	400	0.5241
0.4771	0.1110	500	0.5174
0.4616	0.1331	600	0.5113
0.5277	0.1553	700	0.5073
0.5176	0.1775	800	0.5026
0.5575	0.1997	900	0.4988
0.5059	0.2219	1000	0.4953
0.6007	0.2441	1100	0.4924
0.5123	0.2663	1200	0.4894
0.5547	0.2885	1300	0.4865
0.5183	0.3107	1400	0.4834
0.4759	0.3329	1500	0.4811
0.5157	0.3551	1600	0.4787
0.4501	0.3773	1700	0.4761
0.4594	0.3994	1800	0.4739
0.4579	0.4216	1900	0.4719
0.4539	0.4438	2000	0.4698
0.4225	0.4660	2100	0.4680
0.4594	0.4882	2200	0.4662
0.4248	0.5104	2300	0.4646
0.4287	0.5326	2400	0.4631
0.5521	0.5548	2500	0.4618
0.4582	0.5770	2600	0.4606
0.4871	0.5992	2700	0.4596
0.5356	0.6214	2800	0.4587
0.4403	0.6436	2900	0.4579
0.4056	0.6657	3000	0.4574
0.4131	0.6879	3100	0.4568
0.4544	0.7101	3200	0.4565
0.4971	0.7323	3300	0.4563
0.4663	0.7545	3400	0.4561
0.4744	0.7767	3500	0.4561
0.5264	0.7989	3600	0.4561

Framework versions

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