nguyenthanhdo/ViMath-PAL-deepseek-math-7B-LORA

See axolotl config

axolotl version: 0.4.1

base_model: deepseek-ai/deepseek-math-7b-rl
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false
 
datasets:
  - path: /workspace/axolotl/vinh/PAL/input_output_dsmath.json
    type: input_output
dataset_prepared_path:
val_set_size: 0.05
eval_sample_packing: false
output_dir: /workspace/axolotl/vinh/deepseek-ai_deepseek-math-7b-rl-lora-2024-07-01-15-56-34

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▁sentence｜>

workspace/axolotl/vinh/deepseek-ai_deepseek-math-7b-rl-lora-2024-07-01-15-56-34

This model is a fine-tuned version of deepseek-ai/deepseek-math-7b-rl on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0276

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
0.4197	0.0095	1	0.4128
0.0885	0.1043	11	0.0781
0.0482	0.2086	22	0.0517
0.045	0.3129	33	0.0429
0.0425	0.4172	44	0.0400
0.0411	0.5214	55	0.0379
0.0348	0.6257	66	0.0359
0.0288	0.7300	77	0.0342
0.0339	0.8343	88	0.0331
0.0297	0.9386	99	0.0318
0.0281	1.0429	110	0.0312
0.027	1.1472	121	0.0303
0.023	1.2515	132	0.0298
0.0259	1.3558	143	0.0297
0.0232	1.4600	154	0.0300
0.0203	1.5643	165	0.0291
0.0241	1.6686	176	0.0284
0.0245	1.7729	187	0.0282
0.0222	1.8772	198	0.0277
0.0231	1.9815	209	0.0278
0.0175	2.0858	220	0.0276
0.0165	2.1901	231	0.0281
0.0174	2.2943	242	0.0281
0.021	2.3986	253	0.0279
0.0147	2.5029	264	0.0277
0.0162	2.6072	275	0.0277
0.0206	2.7115	286	0.0276
0.0241	2.8158	297	0.0276
0.0162	2.9201	308	0.0276

Framework versions

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