Hugging Face's logo

NewEden-Forge
/
kyne-14b-lora

PEFT
Safetensors
qwen2
axolotl
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
kyne-14b-lora / README.md
kubernetes-bad's picture
kubernetes-bad
End of training
be6d21f verified
preview code
|
raw
history blame contribute delete
5.39 kB
metadata
library_name: peft
license: apache-2.0
base_model: allura-org/TQ2.5-14B-Sugarquill-v1
tags:
  - axolotl
  - generated_from_trainer
datasets:
  - Nitral-AI/Creative_Writing-ShareGPT
  - Nitral-AI/ARES-ShareGPT
  - NewEden/Claude-Instruct-5K
  - NewEden/OpenCAI-ShareGPT
  - NewEden/PIPPA-Mega-Filtered
  - NewEden/Roleplay-Logs-Sharegpt-Ngram-cleaned
  - Nitral-AI/Creative_Writing-ShareGPT
model-index:
  - name: control-14b-lora
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.6.0

base_model: allura-org/TQ2.5-14B-Sugarquill-v1
strict: false

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_swiglu: true
liger_fused_linear_cross_entropy: false
liger_cross_entropy: false

# Output and HuggingFace
hub_model_id: NewEden/control-14b-lora
hf_use_auth_token: true
hub_strategy: "all_checkpoints"

output_dir: ./outputs/

wandb_project: huggingface
wandb_entity:
wandb_name: Control-14B


chat_template: chatml
group_by_length: false
datasets:
  - path: Nitral-AI/Creative_Writing-ShareGPT
    type: chat_template
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: Nitral-AI/ARES-ShareGPT
    type: chat_template
    chat_template: chatml
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: NewEden/Claude-Instruct-5K
    type: chat_template
    chat_template: chatml
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: NewEden/OpenCAI-ShareGPT
    type: chat_template
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: NewEden/PIPPA-Mega-Filtered
    type: chat_template
    chat_template: chatml
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: NewEden/Roleplay-Logs-Sharegpt-Ngram-cleaned
    type: chat_template
    chat_template: chatml
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn
  - path: Nitral-AI/Creative_Writing-ShareGPT
    type: chat_template
    chat_template: chatml
    roles_to_train: ["gpt"]
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    train_on_eos: turn

#val_set_size: 0.01
#evals_per_epoch: 1
# eval_table_size:
# eval_max_new_tokens: 128
num_epochs: 2

sequence_len: 8192
save_safetensors: true
saves_per_epoch: 2
logging_steps: 1
special_tokens:

# Quantization
bf16: auto
fp16:
tf32: false
## For LoRA
load_in_8bit: false
load_in_4bit: True

# LoRA
peft_use_rslora: true
adapter: qlora
lora_model_dir:
lora_r: 128  
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:

## if oom
# lora_r: 64  
# lora_alpha: 32
# lora_dropout: 0.1


weight_decay: 0.02
max_grad_norm: 1.0

warmup_ratio: 0.05
learning_rate: 0.00002
lr_scheduler: cosine
#lr_scheduler_kwargs:
optimizer: paged_adamw_8bit # usually adamw_torch or paged_adamw_8bit

## Batch Size
gradient_accumulation_steps: 8
micro_batch_size: 1
eval_batch_size: 1

# Optimizations
pad_to_sequence_len: true
sample_packing: true
eval_sample_packing: false
flash_attention: true
xformers_attention:
gradient_checkpointing: "unsloth"
gradient_checkpointing_kwargs:
   use_reentrant: true
local_rank:
early_stopping_patience:
debug:
special_tokens:
  pad_token: <|endoftext|>
  eos_token: <|im_end|>

control-14b-lora

This model is a fine-tuned version of allura-org/TQ2.5-14B-Sugarquill-v1 on the Nitral-AI/Creative_Writing-ShareGPT, the Nitral-AI/ARES-ShareGPT, the NewEden/Claude-Instruct-5K, the NewEden/OpenCAI-ShareGPT, the NewEden/PIPPA-Mega-Filtered, the NewEden/Roleplay-Logs-Sharegpt-Ngram-cleaned and the Nitral-AI/Creative_Writing-ShareGPT datasets.

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: 2e-05
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 8
  • optimizer: Use OptimizerNames.PAGED_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: 43
  • num_epochs: 2

Training results

Framework versions

  • PEFT 0.14.0
  • Transformers 4.47.1
  • Pytorch 2.3.1+cu121
  • Datasets 3.1.0
  • Tokenizers 0.21.0