adapter: qlora
base_model: samoline/a1eb55eb-2c6f-4487-9969-74630487af5c
bf16: auto
chat_template: llama3
dataloader_num_workers: 6
dataset_prepared_path: null
datasets:
- data_files:
  - train_a3d434dc-7bce-43c1-8998-2fa226e7682e.json
  ds_type: json
  format: custom
  path: /workspace/input_data/train_a3d434dc-7bce-43c1-8998-2fa226e7682e.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: null
eval_max_new_tokens: 128
eval_table_size: null
#evals_per_epoch: 1
eval_steps: 20
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
early_stopping_patience: 3
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/be95618a-f904-4bb9-9705-e6cd11e81fbd
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.0003
load_in_4bit: true
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: 8
lora_target_linear: true
lr_scheduler: cosine
#max_steps: 100
micro_batch_size: 2
mlflow_experiment_name: /tmp/train_a3d434dc-7bce-43c1-8998-2fa226e7682e.json
model_type: AutoModelForCausalLM
num_epochs: 4
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: 0
save_steps: 20
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.0002
wandb_entity: null
wandb_mode: online
wandb_name: 9ed9f189-3f66-45a7-9b4b-1516e8b5e7de
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 9ed9f189-3f66-45a7-9b4b-1516e8b5e7de
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

be95618a-f904-4bb9-9705-e6cd11e81fbd

This model is a fine-tuned version of samoline/a1eb55eb-2c6f-4487-9969-74630487af5c on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.8343

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: 0.0003
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 16
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: 4

Training results

Training Loss	Epoch	Step	Validation Loss
1.1819	0.0000	1	0.9233
0.9195	0.0001	20	0.9107
0.8377	0.0003	40	0.8978
0.9675	0.0004	60	0.8916
0.9283	0.0005	80	0.8964
1.1078	0.0006	100	0.8713
0.9021	0.0008	120	0.8634
0.8324	0.0009	140	0.8611
0.9292	0.0010	160	0.8676
1.0229	0.0011	180	0.8560
0.7843	0.0013	200	0.8513
0.8214	0.0014	220	0.8493
1.1907	0.0015	240	0.8502
0.8496	0.0016	260	0.8329
1.129	0.0018	280	0.8396
0.9849	0.0019	300	0.8448
0.9989	0.0020	320	0.8343

Framework versions

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