Training in progress, step 300

all_results.json

@@ -1,13 +1,13 @@
 {
     "epoch": 4.0,
-    "eval_accuracy": 0.9963782696177063,
-    "eval_loss": 0.012937591411173344,
-    "eval_runtime": 48.4901,
-    "eval_samples_per_second": 51.248,
-    "eval_steps_per_second": 6.414,
+    "eval_accuracy": 0.9583333333333334,
+    "eval_loss": 1.039479374885559,
+    "eval_runtime": 1.0974,
+    "eval_samples_per_second": 21.87,
+    "eval_steps_per_second": 2.734,
     "total_flos": 7439897757745152.0,
-    "train_loss": 2.0212895274162292,
-    "train_runtime": 7.7329,
-    "train_samples_per_second": 12.414,
-    "train_steps_per_second": 1.552
+    "train_loss": 1.4209961891174316,
+    "train_runtime": 7.3487,
+    "train_samples_per_second": 13.064,
+    "train_steps_per_second": 1.089
 }

config.json

@@ -9,35 +9,33 @@
   "hidden_dropout_prob": 0.0,
   "hidden_size": 768,
   "id2label": {
-    "0": "carton boxes",
-    "1": "old newspapers",
-    "10": "used juice box",
-    "11": "used plastic bottles labels",
-    "2": "old newspapers with plastic bottles and plastic garbage",
-    "3": "paper boxes and plastic trash bags",
-    "4": "paper waste",
-    "5": "plastic bottles",
-    "6": "plastic bottles and plastic bags with paper boxes",
-    "7": "plastic bottles and plastic garbage",
-    "8": "plastic garbage",
-    "9": "plastic garbage and paper garbage"
+    "0": "aluminium",
+    "1": "batteries",
+    "10": "takeaway cups",
+    "2": "cardboard",
+    "3": "disposable plates",
+    "4": "glass",
+    "5": "hard plastic",
+    "6": "paper",
+    "7": "paper towel",
+    "8": "polystyrene",
+    "9": "soft plastics"
   },
   "image_size": 224,
   "initializer_range": 0.02,
   "intermediate_size": 3072,
   "label2id": {
-    "carton boxes": "0",
-    "old newspapers": "1",
-    "old newspapers with plastic bottles and plastic garbage": "2",
-    "paper boxes and plastic trash bags": "3",
-    "paper waste": "4",
-    "plastic bottles": "5",
-    "plastic bottles and plastic bags with paper boxes": "6",
-    "plastic bottles and plastic garbage": "7",
-    "plastic garbage": "8",
-    "plastic garbage and paper garbage": "9",
-    "used juice box": "10",
-    "used plastic bottles labels": "11"
+    "aluminium": "0",
+    "batteries": "1",
+    "cardboard": "2",
+    "disposable plates": "3",
+    "glass": "4",
+    "hard plastic": "5",
+    "paper": "6",
+    "paper towel": "7",
+    "polystyrene": "8",
+    "soft plastics": "9",
+    "takeaway cups": "10"
   },
   "layer_norm_eps": 1e-12,
   "model_type": "vit",

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 4.0,
-    "eval_accuracy": 0.9963782696177063,
-    "eval_loss": 0.012937591411173344,
-    "eval_runtime": 48.4901,
-    "eval_samples_per_second": 51.248,
-    "eval_steps_per_second": 6.414
+    "eval_accuracy": 0.9583333333333334,
+    "eval_loss": 1.039479374885559,
+    "eval_runtime": 1.0974,
+    "eval_samples_per_second": 21.87,
+    "eval_steps_per_second": 2.734
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17ae2382b9eca4f52fc4d17a9cb546b8971f310d461d0e085613297c710f7ff9
-size 343254736
+oid sha256:ca8ca01f0c031632e2131270776a232480fdf9bfe32ab7ad11e498d1d677a5a9
+size 343251660

runs/Mar29_18-51-27_X5C922065N/events.out.tfevents.1711734830.X5C922065N.77198.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb92060756fa2c4f538ca77c69a721411f79d55d3f3537b2e74a15d4cca51788
+size 405

runs/Mar29_18-51-27_X5C922065N/events.out.tfevents.1711734965.X5C922065N.77198.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b64f0498695854ebda0c5046adb2a9adece839413b4b4c765764e35204ae4c2a
+size 12367

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 4.0,
     "total_flos": 7439897757745152.0,
-    "train_loss": 2.0212895274162292,
-    "train_runtime": 7.7329,
-    "train_samples_per_second": 12.414,
-    "train_steps_per_second": 1.552
+    "train_loss": 1.4209961891174316,
+    "train_runtime": 7.3487,
+    "train_samples_per_second": 13.064,
+    "train_steps_per_second": 1.089
 }

trainer_state.json

@@ -2,36 +2,29 @@
   "best_metric": null,
   "best_model_checkpoint": null,
   "epoch": 4.0,
-  "eval_steps": 1000,
-  "global_step": 12,
+  "eval_steps": 100,
+  "global_step": 8,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
-    {
-      "epoch": 3.33,
-      "grad_norm": 2.8058300018310547,
-      "learning_rate": 3.3333333333333335e-05,
-      "loss": 2.1093,
-      "step": 10
-    },
     {
       "epoch": 4.0,
-      "step": 12,
+      "step": 8,
       "total_flos": 7439897757745152.0,
-      "train_loss": 2.0212895274162292,
-      "train_runtime": 7.7329,
-      "train_samples_per_second": 12.414,
-      "train_steps_per_second": 1.552
+      "train_loss": 1.4209961891174316,
+      "train_runtime": 7.3487,
+      "train_samples_per_second": 13.064,
+      "train_steps_per_second": 1.089
     }
   ],
   "logging_steps": 10,
-  "max_steps": 12,
+  "max_steps": 8,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
-  "save_steps": 1000,
+  "save_steps": 100,
   "total_flos": 7439897757745152.0,
-  "train_batch_size": 8,
+  "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null
 }