Guilherme34/Reasoner-Hermes3-70b-Lora

See axolotl config

axolotl version: 0.4.1

base_model: NousResearch/Hermes-3-Llama-3.1-70B
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

chat_template: llama3
datasets:
  - path: Guilherme34/Reasoner-Dataset-roles-format
    type: chat_template
    chat_template: llama3
    field_messages: messages
    message_field_role: role
    message_field_content: content
    roles:
      system:
        - system
      user:
        - user
      assistant:
        - assistant

dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/lora-out

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 3
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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
s2_attention:

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
   pad_token: <|end_of_text|>

outputs/lora-out

This model is a fine-tuned version of NousResearch/Hermes-3-Llama-3.1-70B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6269

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.4145	0.0833	1	1.3638
1.4133	0.25	3	1.3479
1.1718	0.5	6	1.0840
0.8807	0.75	9	0.8536
0.7696	1.0	12	0.7617
0.5582	1.25	15	0.7075
0.5734	1.5	18	0.6850
0.5593	1.75	21	0.6519
0.5131	2.0	24	0.6315
0.4138	2.25	27	0.6263
0.3607	2.5	30	0.6266
0.3951	2.75	33	0.6272
0.345	3.0	36	0.6269

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1