Upload trained model export (2025-10-13_10-15-42)

README.md

@@ -9,7 +9,7 @@ tags:
 
 # InterVL-HW1
 
-Trained and exported on 2025-10-13_09-20-38.
+Trained and exported on 2025-10-13_10-15-42.
 
 - Backbone: InternVLChatModel
 - AMP dtype: bfloat16

checkpoints/ckpt_epoch0_step0_2025-10-13_10-15-42.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:545b66a066f8aa1a4c44ee42126d699c97935194fcd157ffe7f0d0e90bcba763
+size 6366330923

checkpoints/latest.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a8d346a2d4671049d03732054ae541b214845aca809d199588559a045c133f41
-size 6366333559
+oid sha256:545b66a066f8aa1a4c44ee42126d699c97935194fcd157ffe7f0d0e90bcba763
+size 6366330923

meta.json

@@ -1,23 +1,32 @@
 {
-  "export_time": "2025-10-13_09-20-38",
+  "export_time": "2025-10-13_10-15-42",
   "trainer_cfg": {
     "batch_size": 1,
-    "grad_accum_steps": 4,
+    "grad_accum_steps": 8,
     "num_workers": 2,
     "lr": 2e-05,
     "weight_decay": 0.01,
     "max_epochs": 3,
     "max_grad_norm": 1.0,
-    "warmup_ratio": 0.03,
+    "warmup_ratio": 0.1,
     "log_every": 5,
     "save_dir": "./checkpoints_pt",
     "resume_path": null,
-    "max_seq_len": 2048,
-    "left_ctx": 512,
-    "max_vis_tokens": 32,
+    "max_seq_len": 1792,
+    "left_ctx": 768,
+    "max_vis_tokens": 64,
     "image_token_id": 151643,
     "PAD_ID": 0,
-    "LABEL_PAD": -100
+    "LABEL_PAD": -100,
+    "label_smoothing": 0.1,
+    "train_frames_min": 8,
+    "train_frames_max": 14,
+    "lr_lm": 5e-06,
+    "lr_vision": 0.0001,
+    "lr_connector": 0.0002,
+    "weight_decay_lm": 0.01,
+    "weight_decay_vision": 0.05,
+    "weight_decay_connector": 0.05
   },
   "model_class": "InternVLChatModel",
   "device": "cuda:0",

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f02769347c69ad9319b002c7a9559750b404ec4606980fd0d2ea7ad48d41003d
+oid sha256:62627ee00a36ab893cf2173e77559795db932fae169d3c9993e49426a6dea5d2
 size 2121885120