Model save

README.md

@@ -2,17 +2,13 @@
 license: apache-2.0
 library_name: peft
 tags:
-- alignment-handbook
-- trl
-- sft
-- generated_from_trainer
 - trl
 - sft
 - alignment-handbook
 - generated_from_trainer
 datasets:
-- HuggingFaceH4/ultrachat_200k
-base_model: mistralai/Mistral-7B-v0.1
+- generator
+base_model: alignment-handbook/zephyr-7b-sft-full
 model-index:
 - name: zephyr-7b-pl-qlora
   results: []
@@ -23,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # zephyr-7b-pl-qlora
 
-This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the HuggingFaceH4/ultrachat_200k dataset.
+This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the generator dataset.
 
 ## Model description
 
@@ -47,8 +43,10 @@ The following hyperparameters were used during training:
 - eval_batch_size: 4
 - seed: 42
 - distributed_type: multi-GPU
-- gradient_accumulation_steps: 4
-- total_train_batch_size: 8
+- num_devices: 4
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 16
+- 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

all_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 0.01,
-    "train_loss": 0.6359813857078552,
-    "train_runtime": 2120.9698,
+    "train_loss": 0.6498598456382751,
+    "train_runtime": 1109.6734,
     "train_samples": 207865,
-    "train_samples_per_second": 0.377,
-    "train_steps_per_second": 0.047
+    "train_samples_per_second": 1.442,
+    "train_steps_per_second": 0.09
 }

lora_0/adapter_config.json

@@ -4,7 +4,7 @@
     "base_model_class": "MistralForCausalLM",
     "parent_library": "transformers.models.mistral.modeling_mistral"
   },
-  "base_model_name_or_path": "mistralai/Mistral-7B-v0.1",
+  "base_model_name_or_path": "alignment-handbook/zephyr-7b-sft-full",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,
@@ -23,13 +23,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "gate_proj",
-    "up_proj",
     "q_proj",
-    "v_proj",
+    "gate_proj",
     "o_proj",
     "k_proj",
-    "down_proj"
+    "v_proj",
+    "down_proj",
+    "up_proj"
   ],
   "task_type": null,
   "use_dora": false,

lora_0/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d95350f5041ab4e9166125ad0900a94e5edbb0d39d4fdfef11b30e664c3a0d10
+oid sha256:be25e89d7bcc96fe7078f3f362f8458402b5e2ff4e9c4bd3da4a4348ce5264a7
 size 167832240

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 0.01,
-    "train_loss": 0.6359813857078552,
-    "train_runtime": 2120.9698,
+    "train_loss": 0.6498598456382751,
+    "train_runtime": 1109.6734,
     "train_samples": 207865,
-    "train_samples_per_second": 0.377,
-    "train_steps_per_second": 0.047
+    "train_samples_per_second": 1.442,
+    "train_steps_per_second": 0.09
 }

trainer_state.json

@@ -1,7 +1,7 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.005737810738312797,
+  "epoch": 0.011475127660795226,
   "eval_steps": 500,
   "global_step": 100,
   "is_hyper_param_search": false,
@@ -23,125 +23,125 @@
     {
       "epoch": 0.0,
       "learning_rate": 5e-07,
-      "loss": 0.6921,
+      "loss": 0.6922,
       "step": 10
     },
     {
       "epoch": 0.0,
       "learning_rate": 4.96201938253052e-07,
-      "loss": 0.6886,
+      "loss": 0.689,
       "step": 15
     },
     {
       "epoch": 0.0,
       "learning_rate": 4.849231551964771e-07,
-      "loss": 0.6827,
+      "loss": 0.6841,
       "step": 20
     },
     {
       "epoch": 0.0,
       "learning_rate": 4.6650635094610966e-07,
-      "loss": 0.6745,
+      "loss": 0.6785,
       "step": 25
     },
     {
       "epoch": 0.0,
       "learning_rate": 4.415111107797445e-07,
-      "loss": 0.6645,
+      "loss": 0.6716,
       "step": 30
     },
     {
       "epoch": 0.0,
       "learning_rate": 4.106969024216348e-07,
-      "loss": 0.6549,
+      "loss": 0.6645,
       "step": 35
     },
     {
       "epoch": 0.0,
       "learning_rate": 3.75e-07,
-      "loss": 0.6463,
+      "loss": 0.6583,
       "step": 40
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 3.355050358314172e-07,
-      "loss": 0.6395,
+      "loss": 0.6513,
       "step": 45
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 2.934120444167326e-07,
-      "loss": 0.6305,
+      "loss": 0.6454,
       "step": 50
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 2.5e-07,
-      "loss": 0.6241,
+      "loss": 0.6396,
       "step": 55
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 2.065879555832674e-07,
-      "loss": 0.6155,
+      "loss": 0.635,
       "step": 60
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 1.6449496416858282e-07,
-      "loss": 0.6147,
+      "loss": 0.6302,
       "step": 65
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 1.2500000000000005e-07,
-      "loss": 0.6042,
+      "loss": 0.6281,
       "step": 70
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 8.930309757836516e-08,
-      "loss": 0.6041,
+      "loss": 0.6258,
       "step": 75
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 5.848888922025552e-08,
-      "loss": 0.5989,
+      "loss": 0.6249,
       "step": 80
     },
     {
-      "epoch": 0.0,
+      "epoch": 0.01,
       "learning_rate": 3.349364905389032e-08,
-      "loss": 0.5988,
+      "loss": 0.6218,
       "step": 85
     },
     {
       "epoch": 0.01,
       "learning_rate": 1.507684480352292e-08,
-      "loss": 0.5961,
+      "loss": 0.6206,
       "step": 90
     },
     {
       "epoch": 0.01,
       "learning_rate": 3.798061746947995e-09,
-      "loss": 0.5984,
+      "loss": 0.6223,
       "step": 95
     },
     {
       "epoch": 0.01,
       "learning_rate": 0.0,
-      "loss": 0.5982,
+      "loss": 0.6206,
       "step": 100
     },
     {
       "epoch": 0.01,
       "step": 100,
-      "total_flos": 7.072526927868723e+16,
-      "train_loss": 0.6359813857078552,
-      "train_runtime": 2120.9698,
-      "train_samples_per_second": 0.377,
-      "train_steps_per_second": 0.047
+      "total_flos": 1.4145053855737446e+17,
+      "train_loss": 0.6498598456382751,
+      "train_runtime": 1109.6734,
+      "train_samples_per_second": 1.442,
+      "train_steps_per_second": 0.09
     }
   ],
   "logging_steps": 5,
@@ -149,7 +149,7 @@
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 100,
-  "total_flos": 7.072526927868723e+16,
+  "total_flos": 1.4145053855737446e+17,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null