End of training

README.md

@@ -41,7 +41,7 @@ early_stopping_patience: null
 eval_max_new_tokens: 128
 eval_table_size: null
 evals_per_epoch: 4
-flash_attention: false
+flash_attention: true
 fp16: null
 fsdp: null
 fsdp_config: null
@@ -67,20 +67,20 @@ lr_scheduler: cosine
 max_memory:
   0: 70GiB
 max_steps: 50
-micro_batch_size: 2
+micro_batch_size: 1
 mlflow_experiment_name: /tmp/5daf839d73ce7025_train_data.json
 model_type: AutoModelForCausalLM
 num_epochs: 4
-optimizer: adamw_bnb_8bit
+optimizer: adamw_torch
 output_dir: miner_id_24
 pad_to_sequence_len: true
 resume_from_checkpoint: null
 s2_attention: null
 sample_packing: false
 saves_per_epoch: 4
-sequence_len: 1024
+sequence_len: 2028
 strict: false
-tf32: false
+tf32: true
 tokenizer_type: AutoTokenizer
 train_on_inputs: false
 trust_remote_code: true
@@ -103,7 +103,7 @@ xformers_attention: null
 
 This model is a fine-tuned version of [unsloth/Hermes-3-Llama-3.1-8B](https://huggingface.co/unsloth/Hermes-3-Llama-3.1-8B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: nan
+- Loss: 0.2677
 
 ## Model description
 
@@ -123,12 +123,12 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
-- train_batch_size: 2
-- eval_batch_size: 2
+- train_batch_size: 1
+- eval_batch_size: 1
 - seed: 42
 - gradient_accumulation_steps: 4
-- total_train_batch_size: 8
-- 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: 4
+- 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: 50
@@ -137,19 +137,19 @@ The following hyperparameters were used during training:
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
-| 0.0           | 0.0008 | 1    | nan             |
-| 0.0           | 0.0034 | 4    | nan             |
-| 0.0           | 0.0068 | 8    | nan             |
-| 0.0           | 0.0101 | 12   | nan             |
-| 0.0           | 0.0135 | 16   | nan             |
-| 0.0           | 0.0169 | 20   | nan             |
-| 0.0           | 0.0203 | 24   | nan             |
-| 0.0           | 0.0236 | 28   | nan             |
-| 0.0           | 0.0270 | 32   | nan             |
-| 0.0           | 0.0304 | 36   | nan             |
-| 0.0           | 0.0338 | 40   | nan             |
-| 0.0           | 0.0371 | 44   | nan             |
-| 0.0           | 0.0405 | 48   | nan             |
+| 6.2351        | 0.0004 | 1    | 7.1638          |
+| 7.51          | 0.0017 | 4    | 6.9215          |
+| 4.7822        | 0.0034 | 8    | 4.4631          |
+| 1.5597        | 0.0051 | 12   | 1.5648          |
+| 0.481         | 0.0068 | 16   | 0.6843          |
+| 0.3143        | 0.0084 | 20   | 0.5307          |
+| 0.1861        | 0.0101 | 24   | 0.4394          |
+| 0.1889        | 0.0118 | 28   | 0.4399          |
+| 0.2487        | 0.0135 | 32   | 0.3397          |
+| 0.4158        | 0.0152 | 36   | 0.3151          |
+| 0.3524        | 0.0169 | 40   | 0.2879          |
+| 0.2272        | 0.0186 | 44   | 0.2739          |
+| 0.3537        | 0.0203 | 48   | 0.2677          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:45fe9746a7db9cd15165b811a5331f0dc4b1d8ec660979d5129957450fdbfea6
+oid sha256:1d7c79d0d9671f73d6001c26eafdbb4a800f679b7926a613b593acadba2a3d67
 size 167934026