7b-m-dans-optimizersweeps-repremover-1-adamw_8bit-hi-lr

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.8.0

base_model: Dans-DiscountModels/7b-m-dans-personalityengine-v1.2.1-rc-2
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

trust_remote_code:

# wandb configuration
wandb_project: 7b-m-dans-optimizersweeps
wandb_watch:

wandb_run_id: repremover-1-1-adamw_8bit-hi-lr
wandb_log_model:

# push checkpoints to hub
hub_model_id: Dans-DiscountModels/7b-m-dans-optimizersweeps-repremover-1-adamw_8bit-hi-lr
# how to push checkpoints to hub
# https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.hub_strategy
hub_strategy: "every_save"
# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private datasets
# Required to be true when used in combination with `push_dataset_to_hub`
hf_use_auth_token: true

# where to save the finished model to
output_dir: ./7b-m-dans-optimizersweeps

# where to save the dataset to
dataset_prepared_path: ./7b-m-dans-optimizersweeps-data

save_safetensors: true

# dataset settings (local or huggingface repo)
datasets:
  - path: Dans-DiscountModels/pretokenization-test-3
    ds_type: parquet
    type:

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

load_in_8bit: false
load_in_4bit: false
strict: false

adapter:
lora_model_dir:

val_set_size: 0.01
sequence_len: 8192

sample_packing: false
eval_sample_packing: false

pad_to_sequence_len: true

gradient_checkpointing: true
# gradient_checkpointing_kwargs:
# use_reentrant: false

gradient_accumulation_steps: 1
micro_batch_size: 4

num_epochs: 3

optimizer: adamw_8bit

lr_scheduler: rex
learning_rate: 0.0000003
cosine_min_lr_ratio:

# weight_decay: 0.03
max_grad_norm: 0.001

train_on_inputs: false
group_by_length: true

bf16: true
fp16: false
tf32: false

early_stopping_patience:

resume_from_checkpoint:
auto_resume_from_checkpoints: false

local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_ratio: 0.1

evals_per_epoch: 24
eval_table_size:
eval_max_new_tokens:

saves_per_epoch: 1
save_total_limit: 2

debug: false

deepspeed: deepspeed_configs/zero3_bf16.json

fsdp:
fsdp_config:

special_tokens:

</details><br>

7b-m-dans-optimizersweeps-repremover-1-adamw_8bit-hi-lr

This model is a fine-tuned version of Dans-DiscountModels/7b-m-dans-personalityengine-v1.2.1-rc-2 on the Dans-DiscountModels/pretokenization-test-3 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0812

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: 3e-07
• 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_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: 41
• num_epochs: 3.0

Training results

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1