End of training

README.md

@@ -1,12 +1,12 @@
 ---
 library_name: peft
-license: gemma
-base_model: zake7749/gemma-2-2b-it-chinese-kyara-dpo
+license: cc-by-sa-4.0
+base_model: defog/sqlcoder-7b-2
 tags:
 - axolotl
 - generated_from_trainer
 model-index:
-- name: 4aefa0ce-d88c-4eb6-92b3-06407b441084
+- name: eae8f117-788f-4851-8ae0-704076ca26e8
   results: []
 ---
 
@@ -19,40 +19,39 @@ should probably proofread and complete it, then remove this comment. -->
 axolotl version: `0.4.1`
 ```yaml
 adapter: lora
-base_model: zake7749/gemma-2-2b-it-chinese-kyara-dpo
+base_model: defog/sqlcoder-7b-2
 bf16: auto
 dataset_prepared_path: null
 datasets:
 - data_files:
-  - 8c2dd3c11c63229d_train_data.json
+  - 520602e56e559160_train_data.json
   ds_type: json
   format: custom
-  path: /root/G.O.D-test/core/data/8c2dd3c11c63229d_train_data.json
+  path: /root/G.O.D-test/core/data/520602e56e559160_train_data.json
   type:
-    field_instruction: premise
-    field_output: hypothesis
+    field_instruction: prompt
+    field_output: gold_standard_solution
     format: '{instruction}'
     no_input_format: '{instruction}'
     system_format: '{system}'
     system_prompt: ''
 debug: null
 deepspeed: null
-early_stopping_patience: null
 eval_max_new_tokens: 128
-eval_table_size: null
-evals_per_epoch: 0
+eval_steps: 0
+evals_per_epoch: null
 flash_attention: true
 fp16: null
 fsdp: null
 fsdp_config: null
 gradient_accumulation_steps: 4
 gradient_checkpointing: false
-group_by_length: false
-hub_model_id: souging/4aefa0ce-d88c-4eb6-92b3-06407b441084
+group_by_length: true
+hub_model_id: souging/eae8f117-788f-4851-8ae0-704076ca26e8
 hub_repo: null
 hub_strategy: checkpoint
 hub_token: null
-learning_rate: 0.0002
+learning_rate: 0.000202
 load_in_4bit: false
 load_in_8bit: false
 local_rank: null
@@ -64,44 +63,45 @@ lora_model_dir: null
 lora_r: 32
 lora_target_linear: true
 lr_scheduler: cosine
-max_steps: 400
-micro_batch_size: 3
-mlflow_experiment_name: /tmp/8c2dd3c11c63229d_train_data.json
+max_steps: 500
+micro_batch_size: 2
+mlflow_experiment_name: /tmp/520602e56e559160_train_data.json
 model_type: AutoModelForCausalLM
-num_epochs: 4
+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
-saves_per_epoch: 0
-sequence_len: 2048
+save_steps: 0
+saves_per_epoch: null
+seed: 20
+sequence_len: 1536
+special_tokens:
+  pad_token: </s>
 strict: false
 tf32: false
 tokenizer_type: AutoTokenizer
 train_on_inputs: false
 trust_remote_code: true
-val_set_size: 0.05
 wandb_entity: null
 wandb_mode: online
-wandb_name: 60492357-69eb-4e2c-a118-9f38faabd1d2
+wandb_name: ff2787c9-8dbd-4e47-864f-073c4a88122b
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
-wandb_runid: 60492357-69eb-4e2c-a118-9f38faabd1d2
+wandb_runid: ff2787c9-8dbd-4e47-864f-073c4a88122b
 warmup_steps: 100
-weight_decay: 0.01
+weight_decay: 0.0
 xformers_attention: null
 
 ```
 
 </details><br>
 
-# 4aefa0ce-d88c-4eb6-92b3-06407b441084
+# eae8f117-788f-4851-8ae0-704076ca26e8
 
-This model is a fine-tuned version of [zake7749/gemma-2-2b-it-chinese-kyara-dpo](https://huggingface.co/zake7749/gemma-2-2b-it-chinese-kyara-dpo) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 1.7370
+This model is a fine-tuned version of [defog/sqlcoder-7b-2](https://huggingface.co/defog/sqlcoder-7b-2) on the None dataset.
 
 ## Model description
 
@@ -120,25 +120,22 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 0.0002
-- train_batch_size: 3
-- eval_batch_size: 3
-- seed: 42
+- learning_rate: 0.000202
+- train_batch_size: 2
+- eval_batch_size: 2
+- seed: 20
 - distributed_type: multi-GPU
 - num_devices: 8
 - gradient_accumulation_steps: 4
-- total_train_batch_size: 96
-- total_eval_batch_size: 24
+- total_train_batch_size: 64
+- total_eval_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: 100
-- training_steps: 400
+- training_steps: 500
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss |
-|:-------------:|:------:|:----:|:---------------:|
-| 1.6473        | 0.2399 | 400  | 1.7370          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de1a57dda83020dbf728c45d840ceb687d74825f912baee85ac399e0ad3ecad1
-size 166265274
+oid sha256:3e72e709009678d1415294eb00e7d1c6278129ed98d62dd3870b1e2bd7b17b37
+size 319977674