End of training

README.md

@@ -1,7 +1,7 @@
 ---
 library_name: peft
-license: other
-base_model: Qwen/Qwen1.5-0.5B
+license: llama3
+base_model: NousResearch/Hermes-2-Pro-Llama-3-8B
 tags:
 - axolotl
 - generated_from_trainer
@@ -19,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
 axolotl version: `0.4.1`
 ```yaml
 adapter: lora
-base_model: Qwen/Qwen1.5-0.5B
+base_model: NousResearch/Hermes-2-Pro-Llama-3-8B
 bf16: false
 chat_template: llama3
 dataset_prepared_path: null
@@ -39,23 +39,12 @@ datasets:
     system_prompt: ''
 debug: null
 deepspeed: null
-devices:
-- 0
-- 1
-- 2
-- 3
-- 4
-- 5
-- 6
-- 7
 early_stopping_patience: null
 eval_max_new_tokens: 128
 eval_table_size: null
 evals_per_epoch: 4
-flash_attention: false
+flash_attention: true
 fp16: true
-fsdp: null
-fsdp_config: null
 gradient_accumulation_steps: 4
 gradient_checkpointing: false
 group_by_length: false
@@ -80,8 +69,7 @@ micro_batch_size: 1
 mlflow_experiment_name: /tmp/858b566324e54745_train_data.json
 model_type: AutoModelForCausalLM
 num_epochs: 1
-num_gpus: 8
-optimizer: adamw_bnb_8bit
+optimizer: adamw_torch
 output_dir: miner_id_24
 pad_to_sequence_len: true
 resume_from_checkpoint: null
@@ -112,9 +100,9 @@ xformers_attention: null
 
 # e2b10744-8290-4179-bd2f-9ea76dbe1f32
 
-This model is a fine-tuned version of [Qwen/Qwen1.5-0.5B](https://huggingface.co/Qwen/Qwen1.5-0.5B) on the None dataset.
+This model is a fine-tuned version of [NousResearch/Hermes-2-Pro-Llama-3-8B](https://huggingface.co/NousResearch/Hermes-2-Pro-Llama-3-8B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 6.2907
+- Loss: 8.9346
 
 ## Model description
 
@@ -137,9 +125,12 @@ The following hyperparameters were used during training:
 - train_batch_size: 1
 - eval_batch_size: 1
 - seed: 42
+- distributed_type: multi-GPU
+- num_devices: 2
 - gradient_accumulation_steps: 4
-- total_train_batch_size: 4
-- optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- total_train_batch_size: 8
+- total_eval_batch_size: 2
+- optimizer: Use OptimizerNames.ADAMW_TORCH 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: 10
@@ -149,10 +140,10 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 7.3707        | 0.0000 | 1    | 7.0794          |
-| 4.2049        | 0.0001 | 3    | 7.0794          |
-| 6.3815        | 0.0002 | 6    | 6.9406          |
-| 7.3729        | 0.0002 | 9    | 6.2907          |
+| 10.6832       | 0.0001 | 1    | 9.9402          |
+| 8.955         | 0.0002 | 3    | 9.9402          |
+| 8.8015        | 0.0003 | 6    | 9.9000          |
+| 9.2568        | 0.0005 | 9    | 8.9346          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:304b63ff159de17e7d943af81d65d27fe993f94e7c0981fdb53a9b8c853b2783
-size 30398410
+oid sha256:c638d3d5980ceb5e2a3166b18dfa79735920eeae44823f325610fe56d8e6f150
+size 167934026