Modelpublic

Delta-Vector/Hamanasu-15B-R1-PT

sourceHugging Faceupdated 2y agoView on Hugging Face

3likes12downloads

Model Card

<!DOCTYPE html> <style> html, body { background: black; color: #c9d1d9 !important; font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif; margin: 0; padding: 0; min-height: 100vh; } .markdown-body { color: white; margin: 40px auto; padding: 40px; border-radius: 12px; position: relative; overflow: hidden; }

.markdown-body::after { content: ''; position: absolute; top: 0; left: 0; width: 100%; height: 100%; background: #0c0f18; / background color / pointer-events: none; z-index: -999; }

h1, h2, h3 { background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent; border-bottom: 1px solid #333; padding-bottom: 0.3em; }

div[style="border:2px solid #333"], div[style="border: 2px solid #333"], div[style="border:1px solid #333"], div[style="border: 1px solid #333"] { background: rgba(22, 27, 34, 0.8) !important; border: 2px solid #6e00ff !important; box-shadow: 0 0 15px rgba(110, 0, 255, 0.5); border-radius: 10px; padding: 20px; margin: 20px 0; }

code { background-color: #1a1a1a !important; border-radius: 4px; padding: 0.2em 0.4em; color: #00ffff; }

pre { background-color: #1a1a1a !important; border: 1px solid #333; border-radius: 8px; padding: 16px; }

table { width: 100%; border-collapse: collapse; margin: 20px 0; background: rgba(0,0,0,0.2); table-layout: fixed; color: white; }

th, td { border: 1px solid #333; padding: 12px; text-align: center; color: white; }

th { background: rgba(110, 0, 255, 0.1); }

td:nth-child(1) { width: 1%; white-space: nowrap; }

td:nth-child(2) { width: 100%; }

td > span { display: block; padding: 4px 8px; background: rgba(110, 0, 255, 0.1); border-radius: 4px; transition: all 0.3s ease; }

td > span:hover { background: rgba(110, 0, 255, 0.2); transform: translateY(-1px); }

a { color: #00ffff; text-decoration: none; transition: all 0.3s ease; }

a:hover { color: #6e00ff; text-decoration: none; }

hr { border: 0; height: 1px; background: linear-gradient(90deg, transparent, #333, transparent); margin: 40px 0; }

img { max-width: 100%; border-radius: 10px; }

details summary:hover { color: #00ffff; }

—{ color-scheme: dark !important; }

.prose, .max-w-none, .px-4 { background-color: transparent !important; color: #c9d1d9 !important; } </style> <body> <div class="markdown-body"> <div align="center">

<div style="font-size:1.5em; font-weight:bold; background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent;"> Hamanasu 15B R1 PT </div>

</div>

🌌 Overview

<i>This is the 1st pretrain of Phi-4 with the following: </i>

—NewEden/Orion-LIT

<i>This model has not been instruct tuned, Ablities to converse may be reduced from the original model, If you would like to roleplay, Please use the Instruct version.</i>

</div>

⚔️ Hardware

—4x RTX 3090 GPUs
—Epochs: 1
—Base: Unsloth/phi-4
—Amount of Tokens: 500 Million </div>

Axolotl Config ꒰(˶• ᴗ •˶)꒱

yaml

base_model: unsloth_phi-4
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

  #hub_model_id: NewEden/Phi4-pretrain
  #hub_strategy: "all_checkpoints"
  #push_dataset_to_hub:
  #hf_use_auth_token: true

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

  #plugins:
  #  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

  #cut_cross_entropy: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Mielikki/Erebus-87k
    type: completion
    field: body
  - path: NewEden/Orion-Asstr-Stories-16K
    type: completion
    field: content
shuffle_merged_datasets: true
dataset_prepared_path: prepared_data
val_set_size: 0.0
output_dir: ./phi4-pt-out-r2

sequence_len: 16384
sample_packing: true
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 128
lora_alpha: 16 
lora_dropout: 0.05
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

lora_modules_to_save:
 - embed_tokens
 - lm_head


wandb_project: mag-phi
wandb_entity:
wandb_watch:
wandb_name: attempt-02
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 1
optimizer: paged_ademamix_8bit
lr_scheduler: cosine
learning_rate: 0.00001

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

gradient_checkpointing: unsloth
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 4
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16_cpuoffload_params.json
weight_decay: 0.01
fsdp:
fsdp_config:

</details> </div>

⚡ Credits

<div align="center"> <div style="font-size:0.8em; opacity:0.8;">Made by</div> <div style="font-size:1.2em; font-weight:bold; background: linear-gradient(45deg, #6e00ff, #00ffff); -webkit-background-clip: text; -webkit-text-fill-color: transparent;">Delta-Vector</div> </div>

</div> </body> </html>