See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Vikhrmodels/Vikhr-7B-instruct_0.4
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 5ec7c0c178ba1b04_train_data.json
  ds_type: json
  field: prompt
  path: /workspace/input_data/
  split: train
  type: completion
ddp_find_unused_parameters: false
debug: null
deepspeed: null
early_stopping_patience: null
eps: 1.0e-06
eval_max_new_tokens: 256
eval_table_size: null
evals_per_epoch: 4
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 1
gradient_checkpointing: true
gradient_clipping: 0.5
gradient_normalization: true
greater_is_better: false
group_by_length: false
hub_model_id: CheapsetZero/24ade83e-027e-4383-8762-a809fb955437
learning_rate: 0.00024
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_nan_inf_filter: true
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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 1496
metric_for_best_model: eval_loss
micro_batch_size: 24
min_lr: 4.8e-05
mlflow_experiment_name: /tmp/5ec7c0c178ba1b04_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
reward_model_sampling_temperature: 0.7
s2_attention: null
sample_packing: false
save_total_limit: 3
saves_per_epoch: 4
sequence_len: 1024
skip_nan_gradients: true
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trl:
  adaptive_beta: true
  beta: 0.12
  entropy_coeff: 0.01
  gradient_normalization: true
  kl_monitoring: true
  max_completion_length: 1024
  num_generations: 7
  reward_funcs:
  - rewards_4ae4b631-baf3-48ba-908b-b393d1e81e48.reward_think_answer_format_normalized
  reward_weights:
  - 5.0
  target_kl: 0.01
  use_vllm: false
trust_remote_code: true
use_ema: false
use_peft: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: offline
wandb_name: 4ae4b631-baf3-48ba-908b-b393d1e81e48
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 4ae4b631-baf3-48ba-908b-b393d1e81e48
warmup_steps: 214
weight_decay: 0.01
xformers_attention: null

24ade83e-027e-4383-8762-a809fb955437

This model is a fine-tuned version of Vikhrmodels/Vikhr-7B-instruct_0.4 on the None dataset. It achieves the following results on the evaluation set:

• Loss: nan

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: 0.00024
• train_batch_size: 24
• eval_batch_size: 24
• seed: 42
• 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: 214
• training_steps: 1496

Training results

Training Loss	Epoch	Step	Validation Loss
0.0	0.0016	1	nan
0.0	0.2006	125	nan
0.0	0.4013	250	nan
0.0	0.6019	375	nan
0.0	0.8026	500	nan
0.0	1.0032	625	nan
0.0	1.2039	750	nan
0.0	1.4045	875	nan
0.0	1.6051	1000	nan
4.9718	1.8058	1125	nan
0.0	2.0064	1250	nan
0.0	2.2071	1375	nan

Framework versions

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