nguyenthanhdo/ViMath-Llama-3-8B-LORA

See axolotl config

axolotl version: 0.4.1

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

load_in_8bit: false
load_in_4bit: false
strict: false
 
datasets:
  - path: /workspace/axolotl/vinh/PAL/input_output_llama3.json
    type: input_output
  - path: /workspace/axolotl/vinh/INSTRUCT/input_output_llama3.json
    type: input_output
dataset_prepared_path:
val_set_size: 0.05
eval_sample_packing: false
output_dir: /workspace/axolotl/vinh/NousResearch_Meta-Llama-3-8B-Instruct-lora-2024-06-29-17-22-10

sequence_len: 2048
sample_packing: false
pad_to_sequence_len: false

adapter: lora
lora_model_dir: 
lora_r: 64
lora_alpha: 128
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: 128
micro_batch_size: 1
num_epochs: 3
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 2e-4

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

gradient_checkpointing: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention: 
flash_attention: true
s2_attention:

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
evals_per_epoch: 10
eval_table_size:
eval_max_new_tokens: 512
saves_per_epoch: 2
save_total_limit: 20
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
   pad_token: <|end_of_text|>

workspace/axolotl/vinh/NousResearch_Meta-Llama-3-8B-Instruct-lora-2024-06-29-17-22-10

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: 0.1026

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
• gradient_accumulation_steps: 128
• total_train_batch_size: 128
• 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
0.6579	0.0063	1	0.6361
0.1746	0.1011	16	0.1862
0.1495	0.2023	32	0.1577
0.1288	0.3034	48	0.1459
0.1508	0.4045	64	0.1368
0.1309	0.5056	80	0.1310
0.1179	0.6068	96	0.1283
0.1035	0.7079	112	0.1236
0.1117	0.8090	128	0.1208
0.1126	0.9101	144	0.1188
0.0739	1.0113	160	0.1146
0.0741	1.1124	176	0.1134
0.0746	1.2135	192	0.1137
0.0821	1.3146	208	0.1125
0.0768	1.4158	224	0.1091
0.0627	1.5169	240	0.1069
0.0746	1.6180	256	0.1056
0.0767	1.7191	272	0.1031
0.0775	1.8203	288	0.0996
0.0596	1.9214	304	0.0987
0.0463	2.0225	320	0.0976
0.036	2.1236	336	0.1062
0.0401	2.2248	352	0.1029
0.0462	2.3259	368	0.1039
0.0476	2.4270	384	0.1034
0.0372	2.5281	400	0.1026
0.0377	2.6293	416	0.1026
0.0358	2.7304	432	0.1026
0.0392	2.8315	448	0.1027
0.0384	2.9326	464	0.1026

Framework versions

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