See axolotl config

axolotl version: 0.5.2

adapter: lora
auto_find_batch_size: true
base_model: microsoft/phi-1_5
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - ecf7f142ab6585de_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ecf7f142ab6585de_train_data.json
  type:
    field_instruction: content
    field_output: title
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: /workspace/axolotl/configs/deepspeed_stage2.json
early_stopping: true
early_stopping_patience: 3
eval_max_new_tokens: 128
eval_sample_packing: false
eval_steps: 10
eval_table_size: null
evaluation_strategy: steps
flash_attention: true
fp16: false
gpu_memory_limit: 80GiB
gradient_accumulation_steps: 4
gradient_checkpointing: true
greater_is_better: false
group_by_length: true
hub_model_id: PhoenixB/4198a6da-7d3a-4fec-8edc-86559239004a
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 2e-4
liger_fused_linear_cross_entropy: true
liger_glu_activation: true
liger_layer_norm: true
liger_rms_norm: true
liger_rope: true
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 5
lora_alpha: 16
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_steps: 10000
metric_for_best_model: eval_loss
micro_batch_size: 2
mlflow_experiment_name: /tmp/ecf7f142ab6585de_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch_fused
output_dir: miner_id_24
pad_to_sequence_len: true
plugins:
- axolotl.integrations.liger.LigerPlugin
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 20
save_total_limit: 1
sequence_len: 2048
special_tokens:
  pad_token: <|endoftext|>
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: b662ba6a-4b9d-4ccc-b470-1f1b952f0188
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: b662ba6a-4b9d-4ccc-b470-1f1b952f0188
warmup_steps: 20
weight_decay: 0.0

4198a6da-7d3a-4fec-8edc-86559239004a

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: 1.5252

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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• total_eval_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 20
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0002	1	3.5206
4.053	0.0016	10	3.4192
1.6753	0.0032	20	2.4225
3.2	0.0048	30	2.2037
2.4056	0.0064	40	2.0832
0.8712	0.0081	50	2.0928
2.4406	0.0097	60	1.9572
1.2174	0.0113	70	1.9147
2.7408	0.0129	80	1.8816
1.9455	0.0145	90	1.8502
0.8266	0.0161	100	1.8277
2.2578	0.0177	110	1.7903
1.0678	0.0193	120	1.7695
2.128	0.0209	130	1.7446
1.7646	0.0226	140	1.7429
0.5958	0.0242	150	1.7290
2.3297	0.0258	160	1.6972
0.772	0.0274	170	1.6931
2.4212	0.0290	180	1.6919
1.914	0.0306	190	1.6633
0.8855	0.0322	200	1.6642
2.1254	0.0338	210	1.6393
0.9285	0.0354	220	1.6486
2.3201	0.0371	230	1.6257
1.7027	0.0387	240	1.6169
0.6795	0.0403	250	1.6132
2.0956	0.0419	260	1.6020
0.822	0.0435	270	1.6078
2.4096	0.0451	280	1.5919
1.4415	0.0467	290	1.5861
0.561	0.0483	300	1.5855
2.1833	0.0499	310	1.5695
1.0053	0.0516	320	1.5739
2.2196	0.0532	330	1.5600
1.7903	0.0548	340	1.5633
0.5685	0.0564	350	1.5755
2.1138	0.0580	360	1.5419
0.5894	0.0596	370	1.5549
2.311	0.0612	380	1.5468
1.6716	0.0628	390	1.5319
0.5148	0.0644	400	1.5283
1.9299	0.0661	410	1.5329
0.6835	0.0677	420	1.5275
2.0702	0.0693	430	1.5358
1.5531	0.0709	440	1.5142
0.513	0.0725	450	1.5285
2.1196	0.0741	460	1.5151
0.8019	0.0757	470	1.5252

Framework versions

• PEFT 0.13.2
• Transformers 4.46.3
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3