Edens-Gate/Hamanasu-Magnum-4B-Ckpts

See axolotl config

axolotl version: 0.8.0.dev0

base_model: NewEden/Hamanasu-KTO-V2
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

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

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

val_set_size: 0.01
output_dir: ./outputs/out

adapter:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:

sequence_len: 32768
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

wandb_project: tavbussy
wandb_entity:
wandb_watch:
wandb_name: magnum-attempt-02
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.00001
weight_decay: 0.02
max_grad_norm: 0.2

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

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: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1

debug:
deepspeed: ./deepspeed_configs/zero3_bf16.json
fsdp:
fsdp_config:

special_tokens:
  pad_token: <|finetune_right_pad_id|>

outputs/out

This model is a fine-tuned version of NewEden/Hamanasu-KTO-V2 on the PocketDoc/Dans-Personamaxx-Logs, the anthracite-org/kalo-opus-instruct-22k-no-refusal, the lodrick-the-lafted/kalo-opus-instruct-3k-filtered, the anthracite-org/nopm_claude_writing_fixed, the anthracite-org/kalo_opus_misc_240827, the anthracite-org/kalo_misc_part2, the NewEden/Claude-Instruct-5K and the NewEden/Claude-Instruct-2.7K datasets. It achieves the following results on the evaluation set:

• Loss: 1.2656

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 40
• num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.46	0.0109	1	1.4717
1.3692	0.2514	23	1.3862
1.3288	0.5027	46	1.3275
1.2979	0.7541	69	1.3008
2.4633	1.0109	92	1.2825
1.1345	1.2623	115	1.2762
1.1809	1.5137	138	1.2668
1.145	1.7650	161	1.2586
1.0191	2.0219	184	1.2563
1.0526	2.2732	207	1.2644
1.0341	2.5246	230	1.2593
1.0394	2.7760	253	1.2562
0.9845	3.0328	276	1.2571
0.9583	3.2842	299	1.2655
0.9715	3.5355	322	1.2659
0.9463	3.7869	345	1.2656

Framework versions

• Transformers 4.50.0
• Pytorch 2.5.1+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1