chrisgru/llama33B

See axolotl config

axolotl version: 0.5.0

base_model: meta-llama/Llama-3.2-3B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: true

load_in_8bit: false
load_in_4bit: false
strict: false


datasets:
  - path: chrisgru/ro_wiki_chatml_small
    type: chat_template
    chat_template: llama3
    field_messages: conversations
    message_field_role: from
    message_field_content: value

dataset_prepared_path: /workspace/data/ds_preprocess
val_set_size: 0.01
output_dir: ./data/outputs

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true


#adapter: lora
##lora_model_dir:
#lora_r: 64
#lora_alpha: 16
#lora_dropout: 0.05
#lora_target_linear: true
#lora_fan_in_fan_out:
#lora_modules_to_save:
#  - embed_tokens
#  - lm_head

wandb_project: wiki-llm
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 5e-5


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

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 20
evals_per_epoch: 10
eval_table_size:
saves_per_epoch: 1
#eval_max_new_tokens: 128
save_total_limit: 2
debug:
#deepspeed:
weight_decay: 0.0
# fsdp:
#   - full_shard
#   - auto_wrap
# fsdp_config:
#   fsdp_limit_all_gathers: true
#   fsdp_sync_module_states: true
#   fsdp_offload_params: true
#   fsdp_use_orig_params: false
#   fsdp_cpu_ram_efficient_loading: true
#   fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
#   fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
#   fsdp_state_dict_type: FULL_STATE_DICT
#   fsdp_sharding_strategy: FULL_SHARD
#   fsdp_backward_prefetch: BACKWARD_PRE

seed: 1234
hf_use_auth_token: true
hub_strategy: end
hub_model_id: chrisgru/llama-3.2-3B-rowiki
special_tokens:
  bos_token: "<|begin_of_text|>"
  pad_token: "<|finetune_right_pad_id|>"

llama-3.2-3B-rowiki

This model is a fine-tuned version of meta-llama/Llama-3.2-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5161

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: 5e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 1234
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Use paged_adamw_8bit 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: 20
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.4683	0.0009	1	1.6826
1.7777	0.1001	117	1.6274
1.4701	0.2003	234	1.6031
1.6591	0.3004	351	1.5815
1.664	0.4006	468	1.5587
1.5308	0.5007	585	1.5404
1.3583	0.6009	702	1.5268
1.4297	0.7010	819	1.5198
1.7561	0.8012	936	1.5168
1.6656	0.9013	1053	1.5161

Framework versions

• Transformers 4.46.1
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.3