Update notebooks/qwen-3-0-6b-it-vi-medical-lora.ipynb
Browse files
notebooks/qwen-3-0-6b-it-vi-medical-lora.ipynb
CHANGED
|
@@ -179,8 +179,8 @@
|
|
| 179 |
"import os\n",
|
| 180 |
"api = HfApi()\n",
|
| 181 |
"\n",
|
| 182 |
-
"repo_id = \"
|
| 183 |
-
"save_path = \"Qwen-3-0.6B-
|
| 184 |
"\n",
|
| 185 |
"# Create the directory if it doesn't exist\n",
|
| 186 |
"os.makedirs(save_path, exist_ok=True)\n",
|
|
@@ -519,7 +519,7 @@
|
|
| 519 |
"outputs": [],
|
| 520 |
"source": [
|
| 521 |
"arguments = TrainingArguments( \n",
|
| 522 |
-
" resume_from_checkpoint=\"./Qwen-3-0.6B-
|
| 523 |
" per_device_train_batch_size=8, \n",
|
| 524 |
" per_device_eval_batch_size=8, \n",
|
| 525 |
" gradient_accumulation_steps=4, \n",
|
|
@@ -540,7 +540,7 @@
|
|
| 540 |
" weight_decay=0.01, \n",
|
| 541 |
" lr_scheduler_type=\"linear\", \n",
|
| 542 |
" seed=42, \n",
|
| 543 |
-
" output_dir=\"Qwen-3-0.6B-
|
| 544 |
" report_to=\"none\",\n",
|
| 545 |
" load_best_model_at_end=True, # Load weights with lowest val loss\n",
|
| 546 |
" \t\tgreater_is_better=False,\n",
|
|
@@ -622,7 +622,7 @@
|
|
| 622 |
}
|
| 623 |
],
|
| 624 |
"source": [
|
| 625 |
-
"!ls -a Qwen-3-0.6B-
|
| 626 |
]
|
| 627 |
},
|
| 628 |
{
|
|
@@ -838,8 +838,8 @@
|
|
| 838 |
},
|
| 839 |
"outputs": [],
|
| 840 |
"source": [
|
| 841 |
-
"model.save_pretrained(\"./Qwen-3-0.6B-
|
| 842 |
-
"tokenizer.save_pretrained(\"./Qwen-3-0.6B-
|
| 843 |
]
|
| 844 |
},
|
| 845 |
{
|
|
|
|
| 179 |
"import os\n",
|
| 180 |
"api = HfApi()\n",
|
| 181 |
"\n",
|
| 182 |
+
"repo_id = \"danhtran2mind/Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\"\n",
|
| 183 |
+
"save_path = \"Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\"\n",
|
| 184 |
"\n",
|
| 185 |
"# Create the directory if it doesn't exist\n",
|
| 186 |
"os.makedirs(save_path, exist_ok=True)\n",
|
|
|
|
| 519 |
"outputs": [],
|
| 520 |
"source": [
|
| 521 |
"arguments = TrainingArguments( \n",
|
| 522 |
+
" resume_from_checkpoint=\"./Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\",\n",
|
| 523 |
" per_device_train_batch_size=8, \n",
|
| 524 |
" per_device_eval_batch_size=8, \n",
|
| 525 |
" gradient_accumulation_steps=4, \n",
|
|
|
|
| 540 |
" weight_decay=0.01, \n",
|
| 541 |
" lr_scheduler_type=\"linear\", \n",
|
| 542 |
" seed=42, \n",
|
| 543 |
+
" output_dir=\"Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\", \n",
|
| 544 |
" report_to=\"none\",\n",
|
| 545 |
" load_best_model_at_end=True, # Load weights with lowest val loss\n",
|
| 546 |
" \t\tgreater_is_better=False,\n",
|
|
|
|
| 622 |
}
|
| 623 |
],
|
| 624 |
"source": [
|
| 625 |
+
"!ls -a Qwen-3-0.6B-Instruct-Vi-Medical-LoRA"
|
| 626 |
]
|
| 627 |
},
|
| 628 |
{
|
|
|
|
| 838 |
},
|
| 839 |
"outputs": [],
|
| 840 |
"source": [
|
| 841 |
+
"model.save_pretrained(\"./Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\") # Local saving\n",
|
| 842 |
+
"tokenizer.save_pretrained(\"./Qwen-3-0.6B-Instruct-Vi-Medical-LoRA\")"
|
| 843 |
]
|
| 844 |
},
|
| 845 |
{
|