See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigscience/bloomz-560m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 77f692c8c486c799_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/77f692c8c486c799_train_data.json
  type:
    field_instruction: ru_text
    field_output: text
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: romainnn/beb57acb-e441-465f-abe1-1317402991b0
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0001
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules: null
lr_scheduler: cosine
max_steps: 7344
micro_batch_size: 4
mlflow_experiment_name: /tmp/77f692c8c486c799_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 621695a8-41ca-4ff0-a063-943a82127cac
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 621695a8-41ca-4ff0-a063-943a82127cac
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

beb57acb-e441-465f-abe1-1317402991b0

This model is a fine-tuned version of bigscience/bloomz-560m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5235

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 10
• training_steps: 7344

Training results

Training Loss	Epoch	Step	Validation Loss
18.398	0.0001	1	4.4947
10.5007	0.0080	100	2.7503
9.7118	0.0160	200	2.5660
8.8525	0.0239	300	2.4505
9.8095	0.0319	400	2.3816
7.3997	0.0399	500	2.3277
9.3523	0.0479	600	2.2683
8.5426	0.0559	700	2.2227
8.0084	0.0638	800	2.1798
8.6222	0.0718	900	2.1462
7.5008	0.0798	1000	2.1117
7.9689	0.0878	1100	2.0846
9.268	0.0958	1200	2.0530
8.3926	0.1038	1300	2.0299
8.2008	0.1117	1400	2.0033
7.6241	0.1197	1500	1.9800
7.776	0.1277	1600	1.9504
6.8257	0.1357	1700	1.9285
8.2749	0.1437	1800	1.9164
6.7785	0.1516	1900	1.8930
6.7472	0.1596	2000	1.8768
6.3152	0.1676	2100	1.8570
8.7907	0.1756	2200	1.8444
7.9454	0.1836	2300	1.8242
7.967	0.1915	2400	1.8124
8.0484	0.1995	2500	1.7939
6.3038	0.2075	2600	1.7812
7.576	0.2155	2700	1.7704
8.2322	0.2235	2800	1.7601
6.1708	0.2314	2900	1.7407
6.9358	0.2394	3000	1.7301
6.8908	0.2474	3100	1.7190
7.2734	0.2554	3200	1.7096
6.4909	0.2634	3300	1.6976
8.0803	0.2714	3400	1.6872
7.2216	0.2793	3500	1.6800
8.283	0.2873	3600	1.6634
6.0649	0.2953	3700	1.6537
6.9195	0.3033	3800	1.6501
8.0928	0.3113	3900	1.6415
6.2996	0.3192	4000	1.6337
6.4201	0.3272	4100	1.6233
5.1072	0.3352	4200	1.6175
5.7974	0.3432	4300	1.6093
6.1775	0.3512	4400	1.6034
6.1851	0.3591	4500	1.5980
7.7934	0.3671	4600	1.5897
5.8761	0.3751	4700	1.5822
6.3755	0.3831	4800	1.5780
5.1953	0.3911	4900	1.5733
5.267	0.3991	5000	1.5677
6.3554	0.4070	5100	1.5641
6.4921	0.4150	5200	1.5578
6.3467	0.4230	5300	1.5531
6.4088	0.4310	5400	1.5497
5.6697	0.4390	5500	1.5475
4.553	0.4469	5600	1.5428
6.1535	0.4549	5700	1.5403
5.6773	0.4629	5800	1.5370
6.654	0.4709	5900	1.5338
6.4089	0.4789	6000	1.5327
6.0679	0.4868	6100	1.5311
5.0659	0.4948	6200	1.5294
5.0589	0.5028	6300	1.5276
5.6209	0.5108	6400	1.5271
6.8491	0.5188	6500	1.5251
6.7898	0.5267	6600	1.5252
6.8402	0.5347	6700	1.5237
6.1226	0.5427	6800	1.5237
5.8644	0.5507	6900	1.5227
5.7179	0.5587	7000	1.5228
7.2765	0.5667	7100	1.5223
5.1157	0.5746	7200	1.5220
6.5525	0.5826	7300	1.5235

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1