aosndvklf/pwoefoijlk

See axolotl config

axolotl version: 0.10.0.dev0

base_model: GreenerPastures/Bald-Beaver-8B

load_in_8bit: false
load_in_4bit: false
strict: false

chat_template: qwen3
datasets:
  - path: hardlyworking/EssentialGames
    type: completion

val_set_size: 0.05
output_dir: ./outputs/out
dataset_prepared_path: last_run_prepared
shuffle_merged_datasets: true

  #hub_model_id: hardlyworking/Sugma8B
  #hub_strategy: "all_checkpoints"
  #push_dataset_to_hub:
  #hf_use_auth_token: true

plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
liger_rope: true
liger_rms_norm: true
liger_layer_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: false
cut_cross_entropy: true

sequence_len: 32768
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

wandb_project: joe_Qwen8B
wandb_entity:
wandb_watch:
wandb_name: Qwen8B
wandb_log_model:

evals_per_epoch: 8
eval_table_size:
eval_max_new_tokens: 128

max_grad_norm: 0.0001
gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 1e-5

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
s2_attention:

deepspeed: /alloc/axolotl/deepspeed_configs/zero2.json

warmup_ratio: 0.05
saves_per_epoch: 1
debug:
weight_decay: 0.01
fsdp:
fsdp_config:
special_tokens:
   pad_token:

outputs/out

This model is a fine-tuned version of GreenerPastures/Bald-Beaver-8B on the hardlyworking/EssentialGames dataset. It achieves the following results on the evaluation set:

• Loss: 3.5551

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_bnb_8bit 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: 79
• num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
6.4652	0.0013	1	6.7119
4.5988	0.1252	99	4.6118
3.8234	0.2503	198	3.8138
3.6123	0.3755	297	3.7119
3.6076	0.5006	396	3.6613
3.5705	0.6258	495	3.6305
3.4955	0.7509	594	3.6085
3.4899	0.8761	693	3.5927
3.5015	1.0013	792	3.5788
3.4645	1.1264	891	3.5719
3.4081	1.2516	990	3.5656
3.5246	1.3767	1089	3.5607
3.4124	1.5019	1188	3.5578
3.4495	1.6271	1287	3.5566
3.3886	1.7522	1386	3.5564
3.4256	1.8774	1485	3.5551

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1