Fischerboot/ll3-sophie-new-8ep

See axolotl config

axolotl version: 0.4.1

base_model: Fischerboot/LLama3-Lexi-Aura-3Some-SLERP-SLERP-ql-merge
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

chat_template: llama3
datasets:
  - path: Fischerboot/newnewdataset-sophie
    type: sharegpt
    conversation: llama3
dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./outputs/newdataset-out

adapter: qlora
lora_model_dir:

sequence_len: 128
sample_packing: false
pad_to_sequence_len: true

lora_r: 1024
lora_alpha: 512
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 8
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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

eval_sample_packing: false
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:
  bos_token: "<|begin_of_text|>"
  eos_token: "<|end_of_text|>"
  pad_token: "<|end_of_text|>"

outputs/newdataset-out

This model is a fine-tuned version of Fischerboot/LLama3-Lexi-Aura-3Some-SLERP-SLERP-ql-merge on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2792

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: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 8

Training results

Training Loss	Epoch	Step	Validation Loss
6.3499	0.0034	1	6.0611
1.4549	0.2526	74	1.8669
0.4942	0.5051	148	0.5161
0.5932	0.7577	222	1.2850
0.8581	1.0102	296	0.7266
1.1222	1.2628	370	0.3729
0.4354	1.5154	444	0.4699
0.6122	1.7679	518	0.6806
0.7419	2.0205	592	0.8912
2.7271	2.2730	666	1.2924
0.93	2.5256	740	0.8516
0.7029	2.7782	814	0.5884
0.5606	3.0307	888	0.5291
0.4365	3.2833	962	0.8004
0.2466	3.5358	1036	0.3922
0.6039	3.7884	1110	0.3917
0.1796	4.0410	1184	0.3216
0.3061	4.2935	1258	0.4309
0.7083	4.5461	1332	0.4010
0.3891	4.7986	1406	0.3268
0.331	5.0512	1480	0.3360
0.3014	5.3038	1554	0.2963
0.125	5.5563	1628	0.3096
0.3207	5.8089	1702	0.3020
0.2809	6.0614	1776	0.2849
1.5804	6.3140	1850	0.2801
0.4681	6.5666	1924	0.2826
0.2527	6.8191	1998	0.2793
0.2207	7.0717	2072	0.2787
0.2498	7.3242	2146	0.2799
0.1927	7.5768	2220	0.2798
0.415	7.8294	2294	0.2792

Framework versions

• PEFT 0.11.1
• Transformers 4.41.1
• Pytorch 2.1.2+cu118
• Datasets 2.19.1
• Tokenizers 0.19.1