update model checkpoints

README.md

@@ -1,14 +1,11 @@
 ---
-license: gemma
 library_name: peft
 tags:
 - alignment-handbook
-- trl
-- sft
 - generated_from_trainer
-base_model: google/gemma-7b
 datasets:
 - llama-duo/synth_summarize_dataset_dedup
+base_model: google/gemma-7b
 model-index:
 - name: gemma7b-summarize-gpt4o-128k
   results: []
@@ -21,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the llama-duo/synth_summarize_dataset_dedup dataset.
 It achieves the following results on the evaluation set:
-- Loss: 4.8761
+- Loss: 2.4869
 
 ## Model description
 
@@ -42,43 +39,38 @@ More information needed
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
 - train_batch_size: 4
-- eval_batch_size: 4
+- eval_batch_size: 2
 - seed: 42
 - distributed_type: multi-GPU
-- num_devices: 2
+- num_devices: 8
 - gradient_accumulation_steps: 2
-- total_train_batch_size: 16
-- total_eval_batch_size: 8
+- total_train_batch_size: 64
+- total_eval_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 15
+- num_epochs: 10
 
 ### Training results
 
-| Training Loss | Epoch | Step  | Validation Loss |
-|:-------------:|:-----:|:-----:|:---------------:|
-| 0.8829        | 1.0   | 815   | 2.3801          |
-| 0.8134        | 2.0   | 1630  | 2.3417          |
-| 0.7643        | 3.0   | 2445  | 2.4263          |
-| 0.7211        | 4.0   | 3260  | 2.5582          |
-| 0.6713        | 5.0   | 4075  | 2.7395          |
-| 0.629         | 6.0   | 4890  | 3.0156          |
-| 0.5603        | 7.0   | 5705  | 3.2838          |
-| 0.5081        | 8.0   | 6520  | 3.5777          |
-| 0.4652        | 9.0   | 7335  | 4.0181          |
-| 0.434         | 10.0  | 8150  | 4.3232          |
-| 0.3977        | 11.0  | 8965  | 4.6346          |
-| 0.3878        | 12.0  | 9780  | 4.7574          |
-| 0.3766        | 13.0  | 10595 | 4.8475          |
-| 0.3665        | 14.0  | 11410 | 4.8724          |
-| 0.3658        | 15.0  | 12225 | 4.8761          |
+| Training Loss | Epoch  | Step | Validation Loss |
+|:-------------:|:------:|:----:|:---------------:|
+| 1.1594        | 0.9977 | 219  | 2.6195          |
+| 1.0276        | 2.0    | 439  | 2.4670          |
+| 0.9492        | 2.9977 | 658  | 2.4451          |
+| 0.8751        | 4.0    | 878  | 2.4359          |
+| 0.8477        | 4.9977 | 1097 | 2.4390          |
+| 0.809         | 6.0    | 1317 | 2.4546          |
+| 0.7918        | 6.9977 | 1536 | 2.4592          |
+| 0.7847        | 8.0    | 1756 | 2.4783          |
+| 0.7808        | 8.9977 | 1975 | 2.4889          |
+| 0.7794        | 9.9772 | 2190 | 2.4869          |
 
 
 ### Framework versions
 
 - PEFT 0.10.0
 - Transformers 4.40.0
-- Pytorch 2.2.1+cu121
+- Pytorch 2.1.2+cu121
 - Datasets 2.18.0
 - Tokenizers 0.19.1

adapter_config.json

@@ -10,23 +10,18 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 16,
+  "lora_alpha": 64,
   "lora_dropout": 0.05,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 8,
+  "r": 32,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
-    "down_proj",
-    "k_proj",
     "v_proj",
-    "up_proj",
-    "o_proj",
-    "gate_proj"
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6fd0002e4068a7964c6a73eaeea266c5304d62773841d94dd6bd2a4402d9ffa4
-size 50056096
+oid sha256:598b26d0f7b973fe163ec750abe369f3df5104ffe548f7bd9a95080fa76ea846
+size 25705248

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 4.876120567321777,
-    "eval_runtime": 2.0583,
+    "epoch": 9.977220956719817,
+    "eval_loss": 2.486895799636841,
+    "eval_runtime": 0.2338,
     "eval_samples": 25,
-    "eval_samples_per_second": 4.858,
-    "eval_steps_per_second": 0.972,
-    "total_flos": 9.345237271407755e+18,
-    "train_loss": 0.8459899323898347,
-    "train_runtime": 79658.5706,
-    "train_samples": 119991,
-    "train_samples_per_second": 2.455,
-    "train_steps_per_second": 0.153
+    "eval_samples_per_second": 42.766,
+    "eval_steps_per_second": 4.277,
+    "total_flos": 6.685995643959247e+18,
+    "train_loss": 1.6744230088577967,
+    "train_runtime": 5295.0828,
+    "train_samples": 129221,
+    "train_samples_per_second": 26.519,
+    "train_steps_per_second": 0.414
 }

config.json

@@ -23,9 +23,9 @@
     "_load_in_4bit": true,
     "_load_in_8bit": false,
     "bnb_4bit_compute_dtype": "bfloat16",
-    "bnb_4bit_quant_storage": "bfloat16",
+    "bnb_4bit_quant_storage": "uint8",
     "bnb_4bit_quant_type": "nf4",
-    "bnb_4bit_use_double_quant": true,
+    "bnb_4bit_use_double_quant": false,
     "llm_int8_enable_fp32_cpu_offload": false,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 4.876120567321777,
-    "eval_runtime": 2.0583,
+    "epoch": 9.977220956719817,
+    "eval_loss": 2.486895799636841,
+    "eval_runtime": 0.2338,
     "eval_samples": 25,
-    "eval_samples_per_second": 4.858,
-    "eval_steps_per_second": 0.972
+    "eval_samples_per_second": 42.766,
+    "eval_steps_per_second": 4.277
 }

runs/Jun13_07-22-13_gpu1-2/events.out.tfevents.1718234758.gpu1-2.1174782.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0ba027b8c29cbdae1013ae054b30ce0d300e784aa34d712237e3b8668922e1
+size 101044

runs/Jun13_07-22-13_gpu1-2/events.out.tfevents.1718240054.gpu1-2.1174782.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:468c336c14c5c8dbb22dde0be8ea36b7266b9fdcc5ab175910c28532116ba6f6
+size 359

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 15.0,
-    "total_flos": 9.345237271407755e+18,
-    "train_loss": 0.8459899323898347,
-    "train_runtime": 79658.5706,
-    "train_samples": 119991,
-    "train_samples_per_second": 2.455,
-    "train_steps_per_second": 0.153
+    "epoch": 9.977220956719817,
+    "total_flos": 6.685995643959247e+18,
+    "train_loss": 1.6744230088577967,
+    "train_runtime": 5295.0828,
+    "train_samples": 129221,
+    "train_samples_per_second": 26.519,
+    "train_steps_per_second": 0.414
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a2e6f6a085ea159a9d3d15a7146c7bc6a09e26a3a7db99b2b5fc938ca1950a4c
+oid sha256:a7dddc219ed688da0dbc3783b6e56bb5749c3e2c0e72039146a8794d8a56351c
 size 5176