amphora/plain-lm3-100k-lora

See axolotl config

axolotl version: 0.4.1

base_model: NousResearch/Meta-Llama-3-8B-Instruct
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: true
load_in_4bit: false
strict: false

chat_template: llama3
datasets:
  - path: k2-eval-full.jsonl
    type: chat_template
    chat_template: llama3
    field_messages: message
    message_field_role: role
    message_field_content: content
    roles:
      user:
        - user
      assistant:
        - assistant
val_set_size: 0.01
output_dir: ./outputs/lora-out

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: false

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: fincode
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 32
micro_batch_size: 8
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:
deepspeed: deepspeed_configs/zero1.json

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 2
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/Meta-Llama-3-8B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9072

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 32
• total_train_batch_size: 1024
• total_eval_batch_size: 32
• 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
1.4939	0.0108	1	1.2541
1.2413	0.2485	23	1.0233
1.1686	0.4970	46	0.9725
1.1344	0.7454	69	0.9506
1.0826	0.9939	92	0.9353
1.1045	1.2424	115	0.9285
1.0803	1.4909	138	0.9215
1.0649	1.7394	161	0.9138
1.0492	1.9878	184	0.9110
1.061	2.2363	207	0.9092
1.0397	2.4848	230	0.9081
1.0357	2.7333	253	0.9068
1.025	2.9818	276	0.9072

Framework versions

• PEFT 0.11.1
• Transformers 4.42.4
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1