ToastyPigeon / tq14-gutenberg-test-ws

huggingface.co
Total runs: 16
24-hour runs: 0
7-day runs: 9
30-day runs: 10
Model's Last Updated: January 30 2025

Introduction of tq14-gutenberg-test-ws

Model Details of tq14-gutenberg-test-ws

Built with Axolotl

See axolotl config

axolotl version: 0.6.0

# git clone https://github.com/axolotl-ai-cloud/axolotl
# cd axolotl
# git checkout 3915abee4cba364a83f18cc4b8c2bb571cce3bac
# pip3 install packaging ninja huggingface_hub[cli]
# pip3 install -e '.[flash-attn,deepspeed]'
# apt update && apt install libopenmpi-dev 
# pip install mpi4py
# huggingface-cli login --token $hf_key && wandb login $wandb_key
# python -m axolotl.cli.preprocess new-qwen-14-test.yml
# accelerate launch -m axolotl.cli.train new-qwen-14-test.yml
# python -m axolotl.cli.merge_lora new-qwen-14-test.yml
# huggingface-cli upload ToastyPigeon/tq14-unalign-test train-workspace/merged . --exclude "*.md"

# git clone https://github.com/axolotl-ai-cloud/axolotl && cd axolotl && pip3 install packaging ninja huggingface_hub[cli] && pip3 install -e '.[flash-attn,deepspeed]' && cd .. && huggingface-cli login --token $hf_key && wandb login $wandb_key

# Model
base_model: ToastyPigeon/Qwen2.5-14B-Instruct-1M-Unalign
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false
bf16: true
fp16:
tf32: false
flash_attention: true
special_tokens:

# Output
output_dir: ./train-workspace
hub_model_id: ToastyPigeon/tq14-gutenberg-test-ws
hub_strategy: "all_checkpoints"
resume_from_checkpoint:
saves_per_epoch: 4

# Data
sequence_len: 4096 # fits
min_sample_len: 128
dataset_prepared_path: last_run_prepared
datasets:
  - path: ToastyPigeon/gutenberg-sft
    type: chat_template
    chat_template: chatml
    field_messages: conversations
    message_field_role: from
    message_field_content: value
  - path: sam-paech/gutenberg3-generalfiction-scifi-fantasy-romance-adventure-dpo
    type: chat_template
    chat_template: chatml
    field_messages: chosen
    message_field_role: role
    message_field_content: content
warmup_steps: 20
shuffle_merged_datasets: true
sample_packing: true
pad_to_sequence_len: true

# Batching
num_epochs: 1
gradient_accumulation_steps: 1
micro_batch_size: 8
eval_batch_size: 1

# Evaluation
val_set_size: 50
evals_per_epoch: 10
eval_table_size:
eval_max_new_tokens: 256
eval_sample_packing: false

save_safetensors: true

# WandB
wandb_project: New-Qwen-Test
#wandb_entity:

gradient_checkpointing: 'unsloth'
gradient_checkpointing_kwargs:
  use_reentrant: false

unsloth_cross_entropy_loss: true
#unsloth_lora_mlp: true
#unsloth_lora_qkv: true
#unsloth_lora_o: true

# LoRA
adapter: qlora
lora_model_dir:
lora_r: 32
lora_alpha: 64
lora_dropout: 0.125
lora_target_linear: 
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj
lora_modules_to_save:
#peft_use_dora: true

# Optimizer
optimizer: paged_ademamix_8bit # adamw_8bit
lr_scheduler: cosine
learning_rate: 5e-5
cosine_min_lr_ratio: 0.1
weight_decay: 0.01
max_grad_norm: 1.0

# Misc
train_on_inputs: false
group_by_length: false
early_stopping_patience:
local_rank:
logging_steps: 1
xformers_attention:
debug:
#deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16.json # previously blank
fsdp:
fsdp_config:

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

gc_steps: 10
seed: 69

tq14-gutenberg-test-ws

This model is a fine-tuned version of ToastyPigeon/Qwen2.5-14B-Instruct-1M-Unalign on the ToastyPigeon/gutenberg-sft and the sam-paech/gutenberg3-generalfiction-scifi-fantasy-romance-adventure-dpo datasets. It achieves the following results on the evaluation set:

  • Loss: 2.2556
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: 5e-05
  • train_batch_size: 8
  • eval_batch_size: 1
  • seed: 69
  • optimizer: Use OptimizerNames.PAGED_ADEMAMIX_8BIT and the args are: No additional optimizer arguments
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 20
  • num_epochs: 1
Training results
Training Loss Epoch Step Validation Loss
2.5227 0.0026 1 2.3420
2.389 0.1018 39 2.2861
2.358 0.2037 78 2.2738
2.4781 0.3055 117 2.2676
2.228 0.4073 156 2.2629
2.301 0.5091 195 2.2603
2.2135 0.6110 234 2.2583
2.3153 0.7128 273 2.2572
2.5095 0.8146 312 2.2561
2.3648 0.9164 351 2.2556
Framework versions
  • PEFT 0.14.0
  • Transformers 4.47.1
  • Pytorch 2.5.1+cu124
  • Datasets 3.1.0
  • Tokenizers 0.21.0

Runs of ToastyPigeon tq14-gutenberg-test-ws on huggingface.co

16
Total runs
0
24-hour runs
10
3-day runs
9
7-day runs
10
30-day runs

More Information About tq14-gutenberg-test-ws huggingface.co Model

tq14-gutenberg-test-ws huggingface.co

tq14-gutenberg-test-ws huggingface.co is an AI model on huggingface.co that provides tq14-gutenberg-test-ws's model effect (), which can be used instantly with this ToastyPigeon tq14-gutenberg-test-ws model. huggingface.co supports a free trial of the tq14-gutenberg-test-ws model, and also provides paid use of the tq14-gutenberg-test-ws. Support call tq14-gutenberg-test-ws model through api, including Node.js, Python, http.

tq14-gutenberg-test-ws huggingface.co Url

https://huggingface.co/ToastyPigeon/tq14-gutenberg-test-ws

ToastyPigeon tq14-gutenberg-test-ws online free

tq14-gutenberg-test-ws huggingface.co is an online trial and call api platform, which integrates tq14-gutenberg-test-ws's modeling effects, including api services, and provides a free online trial of tq14-gutenberg-test-ws, you can try tq14-gutenberg-test-ws online for free by clicking the link below.

ToastyPigeon tq14-gutenberg-test-ws online free url in huggingface.co:

https://huggingface.co/ToastyPigeon/tq14-gutenberg-test-ws

tq14-gutenberg-test-ws install

tq14-gutenberg-test-ws is an open source model from GitHub that offers a free installation service, and any user can find tq14-gutenberg-test-ws on GitHub to install. At the same time, huggingface.co provides the effect of tq14-gutenberg-test-ws install, users can directly use tq14-gutenberg-test-ws installed effect in huggingface.co for debugging and trial. It also supports api for free installation.

tq14-gutenberg-test-ws install url in huggingface.co:

https://huggingface.co/ToastyPigeon/tq14-gutenberg-test-ws

Url of tq14-gutenberg-test-ws

tq14-gutenberg-test-ws huggingface.co Url

Provider of tq14-gutenberg-test-ws huggingface.co

ToastyPigeon
ORGANIZATIONS

Other API from ToastyPigeon