Model save

README.md

@@ -2,7 +2,6 @@
 license: apache-2.0
 library_name: peft
 tags:
-- axolotl
 - generated_from_trainer
 base_model: mistralai/Mistral-7B-Instruct-v0.2
 model-index:
@@ -37,7 +36,7 @@ datasets:
 
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.1
-output_dir: ../../text-generation-webui/loras/mistral-instruct-raw-format-v2-more-positive-new-tokens-inst
+output_dir: ../../text-generation-webui/loras/mistral-instruct-raw-format-v2-more-positive-inst
 
 adapter: lora
 lora_model_dir:
@@ -52,14 +51,6 @@ lora_dropout: 0.1
 lora_target_linear: true
 lora_fan_in_fan_out:
 
-lora_modules_to_save:
- - embed_tokens
- - lm_head
-
-tokens:
-  function_token: "[f]"
-  conversation_token: "[c]"
-
 hub_model_id: liuylhf/mistral-lora
 
 wandb_project: function-call
@@ -68,7 +59,7 @@ wandb_log_model: end
 
 gradient_accumulation_steps: 4
 micro_batch_size: 2
-num_epochs: 1
+num_epochs: 0.5
 optimizer: paged_adamw_8bit
 lr_scheduler: cosine
 learning_rate: 0.001
@@ -95,8 +86,8 @@ flash_attention: true
 
 warmup_steps: 10
 # evals_per_epoch: 20
-# eval_steps: 0.1
-save_steps: 0.1
+eval_steps: 0.2
+save_steps: 0.2
 eval_table_size:
 eval_max_new_tokens: 256
 # saves_per_epoch: 1
@@ -111,9 +102,7 @@ fsdp_config:
 
 # mistral-lora
 
-This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on the None dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.2163
+This model is a fine-tuned version of [mistralai/Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2) on an unknown dataset.
 
 ## Model description
 
@@ -144,14 +133,7 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_steps: 10
-- num_epochs: 1
-
-### Training results
-
-| Training Loss | Epoch | Step | Validation Loss |
-|:-------------:|:-----:|:----:|:---------------:|
-| 0.149         | 1.0   | 304  | 0.2163          |
-
+- num_epochs: 0.5
 
 ### Framework versions