Hamanasu 32B
## π Overview
This model is the Instruct tuned version of Hamanasu-QwQ-V1, This model removes the reasoning gimmick of QwQ.
Read more about the model's training on my blog : https://openai-sucks.bearblog.dev/.
The model has dry but good prose and stays terse, All thanks to Ruka-Hamanasu for funding the train.
### π Quantizations
| Type | Link |
|:---:|:---:|
| `GGUF` | https://huggingface.co/Delta-Vector/Hamanasu-QwQ-V1.5-Instruct-gguf> |
| `EXL2` | https://huggingface.co/Delta-Vector/Hamanasu-QwQ-V1.5-Instruct-EXL2 |
### βοΈ Hardware
- 8x H100s
- Epochs: 2
- Base: `Delta-Vector/Hamanasu-32B-V1-QwQ`
- Amount of Tokens: 60M
## π° Prompting
This model uses ChatML formatting
```python
<|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.
```
## π² Recommended Sampler Preset
```python
temperature: 1.1
min_p: 0.1
```
## Axolotl Config κ°(ΛΆβ’ α΄ β’ΛΆ)κ±
```yaml
base_model: NewEden/Hamanasu-32B-V1
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
hub_model_id: NewEden/Hamanasu-FFT-Instruct
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
load_in_8bit: false
load_in_4bit: false
strict: false
datasets:
- path: NewEden/Hydrus-R1-Thinking-Sharegpt
type: dan-chat-advanced
- path: PocketDoc/Dans-MemoryCore-CoreCurriculum-Small
type: dan-chat-advanced
- path: Nitral-AI/ARES-ShareGPT
type: dan-chat-advanced
- path: NewEden/Hydrus-HelpSteer2
type: dan-chat-advanced
- path: PocketDoc/Dans-Codemaxx-CodeFeedback-Conversations
type: dan-chat-advanced
- path: PocketDoc/Dans-Toolmaxx-Agent
type: dan-chat-advanced
- path: PocketDoc/Dans-Assistantmaxx-Tulu3-IF
type: dan-chat-advanced
- path: NewEden/Hydrus-SonnetOrca
type: dan-chat-advanced
- path: NewEden/Hydrus-Chat_error-Pure-Dove-sharegpt
type: dan-chat-advanced
- path: NewEden/No_Robots-R1-Filtered
type: dan-chat-advanced
- path: NewEden/GSM8K-R1-filtered
type: dan-chat-advanced
- path: NewEden/Hydrus_Anthropic_hh_harmful-sharegpt
type: dan-chat-advanced
- path: NewEden/Hydrus-Instruct-SmolTalk
type: dan-chat-advanced
- path: PocketDoc/Dans-Logicmaxx-Skunkworks
type: dan-chat-advanced
- path: PocketDoc/Dans-Logicmaxx-SAT-AP
type: dan-chat-advanced
- path: PocketDoc/Dans-Toolmaxx-ShellCommands
type: dan-chat-advanced
- path: PocketDoc/Dans-Taskmaxx-Edit
type: dan-chat-advanced
dataset_prepared_path: prepared_data
val_set_size: 0.0
output_dir: ./qwq-inst
sequence_len: 32768
sample_packing: true
pad_to_sequence_len: true
wandb_project: qwq
wandb_entity:
wandb_watch:
wandb_name: instruct-attempt-03
wandb_log_model:
gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 5e-6
train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
warmup_steps: 40
evals_per_epoch:
eval_table_size:
eval_max_new_tokens:
saves_per_epoch: 2
debug:
deepspeed: deepspeed_configs/zero3_bf16.json
weight_decay: 0.02
fsdp:
fsdp_config:
special_tokens:
```
## β‘ Credits
---
Made by
Delta-Vector