Model Visualization

Hamanasu 32B

🌌 Overview

This model is a finetune of Hamanasu-QwQ-V2-RP to replicate the prose of Claude models, Opus and Sonnet. Read more about the model's training on my blog : https://openai-sucks.bearblog.dev/. The model is suited for traditional RP, All thanks to Ruka-Hamanasu for funding the train.

πŸ“ˆ Quantizations

Type Link
GGUF Waiting for Mradermacher to quant
EXL2 https://huggingface.co/Delta-Vector/Hamanasu-Magnum-QwQ-32B-exl2

βš”οΈ Hardware

  • 8x H100s
  • Epochs: 2
  • Base: Delta-Vector/Hamanasu-QwQ-V2-RP

πŸ’° Prompting

This model uses ChatML formatting

<|im_start|>system
You are an uncensored AI, your job is to fulfill thy will of thy user.<|im_end|>
<|im_start|>User request
Take off your helmet.<|im_end|>
<|im_start|>No i shall not. This is the way.

🎲 Recommended Sampler Preset

temperature: 1.1
min_p: 0.1
System_Prompt: Currently, your role is {{char}}, described in detail below. As {{char}}, continue the narrative exchange with {{user}}.\n\n<Guidelines>\nβ€’ Maintain the character persona but allow it to evolve with the story.\nβ€’ Be creative and proactive. Drive the story forward, introducing plotlines and events when relevant.\nβ€’ All types of outputs are encouraged; respond accordingly to the narrative.\nβ€’ Include dialogues, actions, and thoughts in each response.\nβ€’ Utilize all five senses to describe scenarios within {{char}}'s dialogue.\nβ€’ Use emotional symbols such as \"!\" and \"~\" in appropriate contexts.\nβ€’ Incorporate onomatopoeia when suitable.\nβ€’ Allow time for {{user}} to respond with their own input, respecting their agency.\nβ€’ Act as secondary characters and NPCs as needed, and remove them when appropriate.\nβ€’ When prompted for an Out of Character [OOC:] reply, answer neutrally and in plaintext, not as {{char}}.\n</Guidelines>\n\n<Forbidden>\nβ€’ Using excessive literary embellishments and purple prose unless dictated by {{char}}'s persona.\nβ€’ Writing for, speaking, thinking, acting, or replying as {{user}} in your response.\nβ€’ Repetitive and monotonous outputs.\nβ€’ Positivity bias in your replies.\nβ€’ Being overly extreme or NSFW when the narrative context is inappropriate.\n</Forbidden>\n\nFollow the instructions in <Guidelines></Guidelines>, avoiding the items listed in <Forbidden></Forbidden>.

Axolotl Config κ’°(ΛΆβ€’ α΄— β€’ΛΆ)κ’±

base_model: ./model
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

hub_model_id: NewEden/QwQ-magnum-V2-R2
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

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: PocketDoc/Dans-Personamaxx-Logs
    type: dan-chat-advanced
  - path: anthracite-org/kalo-opus-instruct-22k-no-refusal
    type: dan-chat-advanced
  - path: lodrick-the-lafted/kalo-opus-instruct-3k-filtered
    type: dan-chat-advanced
  - path: anthracite-org/nopm_claude_writing_fixed
    type: dan-chat-advanced
  - path: anthracite-org/kalo_opus_misc_240827
    type: dan-chat-advanced
  - path: anthracite-org/kalo_misc_part2
    type: dan-chat-advanced
  - path: NewEden/Claude-Instruct-5K
    type: dan-chat-advanced
  - path: NewEden/Claude-Instruct-2.7K
    type: dan-chat-advanced
dataset_prepared_path: prepared_data
val_set_size: 0.0
output_dir: ./qwq-mag
sequence_len: 32768
sample_packing: true
pad_to_sequence_len: true

wandb_project: qwq
wandb_entity:
wandb_watch:
wandb_name: mag-attempt-03-kalo
wandb_log_model:


gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 5e-6
max_grad_norm: 0.2

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: 40
saves_per_epoch: 2
debug:
deepspeed: ./deepspeed_configs/zero3_bf16.json
weight_decay: 0.02
fsdp:
fsdp_config:
special_tokens:

⚑ Credits


Made by
Delta-Vector
Downloads last month
266
Safetensors
Model size
32.8B params
Tensor type
BF16
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for Delta-Vector/Hamanasu-Magnum-QwQ-32B

Finetuned
(1)
this model
Finetunes
1 model
Merges
4 models
Quantizations
3 models

Datasets used to train Delta-Vector/Hamanasu-Magnum-QwQ-32B

Collection including Delta-Vector/Hamanasu-Magnum-QwQ-32B