aissatoubalde
commited on
Commit
β’
e9cc5c3
1
Parent(s):
a31ac62
Training in progress, epoch 0
Browse files- .ipynb_checkpoints/phi-2-custom-checkpoint.ipynb +0 -0
- adapter_config.json +3 -2
- adapter_model.safetensors +2 -2
- fine-tuned-phi2.ipynb +0 -0
- phi-2-custom.ipynb +609 -66
- training_args.bin +1 -1
.ipynb_checkpoints/phi-2-custom-checkpoint.ipynb
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
adapter_config.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"alpha_pattern": {},
|
3 |
"auto_mapping": null,
|
4 |
-
"base_model_name_or_path":
|
5 |
"bias": "none",
|
6 |
"fan_in_fan_out": false,
|
7 |
"inference_mode": true,
|
@@ -19,8 +19,9 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
|
|
|
|
22 |
"v_proj",
|
23 |
-
"k_proj",
|
24 |
"q_proj"
|
25 |
],
|
26 |
"task_type": "CAUSAL_LM",
|
|
|
1 |
{
|
2 |
"alpha_pattern": {},
|
3 |
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "microsoft/phi-2",
|
5 |
"bias": "none",
|
6 |
"fan_in_fan_out": false,
|
7 |
"inference_mode": true,
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
+
"fc2",
|
23 |
+
"fc1",
|
24 |
"v_proj",
|
|
|
25 |
"q_proj"
|
26 |
],
|
27 |
"task_type": "CAUSAL_LM",
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b3dc96a838978069f70bac32e97b1ef192924aca0ef3b8d35e67035a4223c6e9
|
3 |
+
size 73433680
|
fine-tuned-phi2.ipynb
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
phi-2-custom.ipynb
CHANGED
@@ -547,7 +547,7 @@
|
|
547 |
},
|
548 |
{
|
549 |
"cell_type": "code",
|
550 |
-
"execution_count":
|
551 |
"id": "18d5599f-992d-4d8e-a90c-4d43774be473",
|
552 |
"metadata": {},
|
553 |
"outputs": [
|
@@ -555,7 +555,7 @@
|
|
555 |
"name": "stdout",
|
556 |
"output_type": "stream",
|
557 |
"text": [
|
558 |
-
"trainable params:
|
559 |
]
|
560 |
}
|
561 |
],
|
@@ -563,7 +563,7 @@
|
|
563 |
"config = LoraConfig(\n",
|
564 |
" r=16,\n",
|
565 |
" lora_alpha=16,\n",
|
566 |
-
" target_modules=[\"q_proj\",\"k_proj\",\"v_proj\"],\n",
|
567 |
" lora_dropout=0.05,\n",
|
568 |
" bias=\"none\",\n",
|
569 |
" task_type=\"CAUSAL_LM\"\n",
|
@@ -575,7 +575,7 @@
|
|
575 |
},
|
576 |
{
|
577 |
"cell_type": "code",
|
578 |
-
"execution_count":
|
579 |
"id": "baeee903-3dce-48b2-93c3-7a697d8c6daf",
|
580 |
"metadata": {},
|
581 |
"outputs": [],
|
@@ -587,10 +587,20 @@
|
|
587 |
},
|
588 |
{
|
589 |
"cell_type": "code",
|
590 |
-
"execution_count":
|
591 |
"id": "28a9b24a-a822-4fcb-96b3-d77b7ea30a5f",
|
592 |
"metadata": {},
|
593 |
"outputs": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
594 |
{
|
595 |
"name": "stdout",
|
596 |
"output_type": "stream",
|
@@ -609,7 +619,7 @@
|
|
609 |
},
|
610 |
{
|
611 |
"cell_type": "code",
|
612 |
-
"execution_count":
|
613 |
"id": "1ee7fd2a-38e4-4f23-a978-0bdeeda64d8b",
|
614 |
"metadata": {},
|
615 |
"outputs": [],
|
@@ -624,7 +634,7 @@
|
|
624 |
},
|
625 |
{
|
626 |
"cell_type": "code",
|
627 |
-
"execution_count":
|
628 |
"id": "e84c29e2-843e-42c2-8c0f-324d392e671c",
|
629 |
"metadata": {},
|
630 |
"outputs": [
|
@@ -632,7 +642,7 @@
|
|
632 |
"name": "stderr",
|
633 |
"output_type": "stream",
|
634 |
"text": [
|
635 |
-
"Tokenizing data: 100%|ββββββββββ| 16412/16412 [00:02<00:00,
|
636 |
]
|
637 |
},
|
638 |
{
|
@@ -644,7 +654,7 @@
|
|
644 |
"})"
|
645 |
]
|
646 |
},
|
647 |
-
"execution_count":
|
648 |
"metadata": {},
|
649 |
"output_type": "execute_result"
|
650 |
}
|
@@ -659,7 +669,7 @@
|
|
659 |
},
|
660 |
{
|
661 |
"cell_type": "code",
|
662 |
-
"execution_count":
|
663 |
"id": "ac968254-5338-49df-950d-222b82647407",
|
664 |
"metadata": {},
|
665 |
"outputs": [],
|
@@ -680,7 +690,7 @@
|
|
680 |
},
|
681 |
{
|
682 |
"cell_type": "code",
|
683 |
-
"execution_count":
|
684 |
"id": "729df2d0-0890-4ac4-adf3-c167a6e9669d",
|
685 |
"metadata": {},
|
686 |
"outputs": [
|
@@ -719,7 +729,7 @@
|
|
719 |
},
|
720 |
{
|
721 |
"cell_type": "code",
|
722 |
-
"execution_count":
|
723 |
"id": "525f81a4-eb92-466e-bb9c-cd63122231ab",
|
724 |
"metadata": {},
|
725 |
"outputs": [
|
@@ -737,11 +747,8 @@
|
|
737 |
"name": "stdout",
|
738 |
"output_type": "stream",
|
739 |
"text": [
|
740 |
-
"Requirement already satisfied: jupyterlab in /usr/local/lib/python3.11/site-packages (4.1.
|
741 |
-
"
|
742 |
-
" Downloading jupyterlab-4.1.1-py3-none-any.whl.metadata (15 kB)\n",
|
743 |
-
"Collecting jupyterlab-git\n",
|
744 |
-
" Downloading jupyterlab_git-0.50.0-py3-none-any.whl.metadata (31 kB)\n",
|
745 |
"Requirement already satisfied: async-lru>=1.0.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (2.0.4)\n",
|
746 |
"Requirement already satisfied: httpx>=0.25.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (0.26.0)\n",
|
747 |
"Requirement already satisfied: ipykernel in /usr/local/lib/python3.11/site-packages (from jupyterlab) (6.29.1)\n",
|
@@ -754,8 +761,7 @@
|
|
754 |
"Requirement already satisfied: packaging in /usr/local/lib/python3.11/site-packages (from jupyterlab) (23.2)\n",
|
755 |
"Requirement already satisfied: tornado>=6.2.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (6.4)\n",
|
756 |
"Requirement already satisfied: traitlets in /usr/local/lib/python3.11/site-packages (from jupyterlab) (5.14.1)\n",
|
757 |
-
"
|
758 |
-
" Downloading nbdime-4.0.1-py3-none-any.whl.metadata (9.5 kB)\n",
|
759 |
"Requirement already satisfied: nbformat in /usr/local/lib/python3.11/site-packages (from jupyterlab-git) (5.9.2)\n",
|
760 |
"Requirement already satisfied: pexpect in /usr/local/lib/python3.11/site-packages (from jupyterlab-git) (4.9.0)\n",
|
761 |
"Requirement already satisfied: anyio in /usr/local/lib/python3.11/site-packages (from httpx>=0.25.0->jupyterlab) (4.2.0)\n",
|
@@ -781,14 +787,10 @@
|
|
781 |
"Requirement already satisfied: json5>=0.9.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (0.9.14)\n",
|
782 |
"Requirement already satisfied: jsonschema>=4.18.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (4.21.1)\n",
|
783 |
"Requirement already satisfied: requests>=2.31 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (2.31.0)\n",
|
784 |
-
"
|
785 |
-
"
|
786 |
-
"
|
787 |
-
"
|
788 |
-
"Collecting jupyter-server-mathjax>=0.2.2 (from nbdime~=4.0.1->jupyterlab-git)\n",
|
789 |
-
" Downloading jupyter_server_mathjax-0.2.6-py3-none-any.whl (3.1 MB)\n",
|
790 |
-
"\u001b[2K \u001b[90mββββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m3.1/3.1 MB\u001b[0m \u001b[31m265.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
791 |
-
"\u001b[?25hRequirement already satisfied: pygments in /usr/local/lib/python3.11/site-packages (from nbdime~=4.0.1->jupyterlab-git) (2.17.2)\n",
|
792 |
"Requirement already satisfied: fastjsonschema in /usr/local/lib/python3.11/site-packages (from nbformat->jupyterlab-git) (2.19.1)\n",
|
793 |
"Requirement already satisfied: comm>=0.1.1 in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (0.2.1)\n",
|
794 |
"Requirement already satisfied: debugpy>=1.6.5 in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (1.8.0)\n",
|
@@ -797,8 +799,7 @@
|
|
797 |
"Requirement already satisfied: nest-asyncio in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (1.6.0)\n",
|
798 |
"Requirement already satisfied: psutil in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (5.9.8)\n",
|
799 |
"Requirement already satisfied: ptyprocess>=0.5 in /usr/local/lib/python3.11/site-packages (from pexpect->jupyterlab-git) (0.7.0)\n",
|
800 |
-
"
|
801 |
-
" Downloading gitdb-4.0.11-py3-none-any.whl.metadata (1.2 kB)\n",
|
802 |
"Requirement already satisfied: decorator in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (5.1.1)\n",
|
803 |
"Requirement already satisfied: jedi>=0.16 in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (0.19.1)\n",
|
804 |
"Requirement already satisfied: prompt-toolkit<3.1.0,>=3.0.41 in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (3.0.43)\n",
|
@@ -825,8 +826,7 @@
|
|
825 |
"Requirement already satisfied: argon2-cffi-bindings in /usr/local/lib/python3.11/site-packages (from argon2-cffi->jupyter-server<3,>=2.4.0->jupyterlab) (21.2.0)\n",
|
826 |
"Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.11/site-packages (from bleach!=5.0.0->nbconvert>=6.4.4->jupyter-server<3,>=2.4.0->jupyterlab) (1.16.0)\n",
|
827 |
"Requirement already satisfied: webencodings in /usr/local/lib/python3.11/site-packages (from bleach!=5.0.0->nbconvert>=6.4.4->jupyter-server<3,>=2.4.0->jupyterlab) (0.5.1)\n",
|
828 |
-
"
|
829 |
-
" Downloading smmap-5.0.1-py3-none-any.whl.metadata (4.3 kB)\n",
|
830 |
"Requirement already satisfied: parso<0.9.0,>=0.8.3 in /usr/local/lib/python3.11/site-packages (from jedi>=0.16->ipython>=7.23.1->ipykernel->jupyterlab) (0.8.3)\n",
|
831 |
"Requirement already satisfied: fqdn in /usr/local/lib/python3.11/site-packages (from jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (1.5.1)\n",
|
832 |
"Requirement already satisfied: isoduration in /usr/local/lib/python3.11/site-packages (from jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (20.11.0)\n",
|
@@ -842,23 +842,6 @@
|
|
842 |
"Requirement already satisfied: pycparser in /usr/local/lib/python3.11/site-packages (from cffi>=1.0.1->argon2-cffi-bindings->argon2-cffi->jupyter-server<3,>=2.4.0->jupyterlab) (2.21)\n",
|
843 |
"Requirement already satisfied: arrow>=0.15.0 in /usr/local/lib/python3.11/site-packages (from isoduration->jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (1.3.0)\n",
|
844 |
"Requirement already satisfied: types-python-dateutil>=2.8.10 in /usr/local/lib/python3.11/site-packages (from arrow>=0.15.0->isoduration->jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (2.8.19.20240106)\n",
|
845 |
-
"Downloading jupyterlab-4.1.1-py3-none-any.whl (11.4 MB)\n",
|
846 |
-
"\u001b[2K \u001b[90mββββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m11.4/11.4 MB\u001b[0m \u001b[31m434.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
847 |
-
"\u001b[?25hDownloading jupyterlab_git-0.50.0-py3-none-any.whl (1.2 MB)\n",
|
848 |
-
"\u001b[2K \u001b[90mββββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m1.2/1.2 MB\u001b[0m \u001b[31m546.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
849 |
-
"\u001b[?25hDownloading nbdime-4.0.1-py3-none-any.whl (5.9 MB)\n",
|
850 |
-
"\u001b[2K \u001b[90mββββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m5.9/5.9 MB\u001b[0m \u001b[31m279.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
851 |
-
"\u001b[?25hDownloading GitPython-3.1.41-py3-none-any.whl (196 kB)\n",
|
852 |
-
"\u001b[2K \u001b[90mβββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m196.4/196.4 kB\u001b[0m \u001b[31m474.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
853 |
-
"\u001b[?25hDownloading gitdb-4.0.11-py3-none-any.whl (62 kB)\n",
|
854 |
-
"\u001b[2K \u001b[90mββββββββββββββββββββββββββββββββββββββββ\u001b[0m \u001b[32m62.7/62.7 kB\u001b[0m \u001b[31m406.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
|
855 |
-
"\u001b[?25hDownloading smmap-5.0.1-py3-none-any.whl (24 kB)\n",
|
856 |
-
"Installing collected packages: smmap, colorama, gitdb, gitpython, jupyter-server-mathjax, nbdime, jupyterlab, jupyterlab-git\n",
|
857 |
-
" Attempting uninstall: jupyterlab\n",
|
858 |
-
" Found existing installation: jupyterlab 4.1.0\n",
|
859 |
-
" Uninstalling jupyterlab-4.1.0:\n",
|
860 |
-
" Successfully uninstalled jupyterlab-4.1.0\n",
|
861 |
-
"Successfully installed colorama-0.4.6 gitdb-4.0.11 gitpython-3.1.41 jupyter-server-mathjax-0.2.6 jupyterlab-4.1.1 jupyterlab-git-0.50.0 nbdime-4.0.1 smmap-5.0.1\n",
|
862 |
"Note: you may need to restart the kernel to use updated packages.\n"
|
863 |
]
|
864 |
}
|
@@ -914,7 +897,9 @@
|
|
914 |
"cell_type": "code",
|
915 |
"execution_count": null,
|
916 |
"id": "3bf553b6-b26c-49c3-9407-74c8d53a395e",
|
917 |
-
"metadata": {
|
|
|
|
|
918 |
"outputs": [
|
919 |
{
|
920 |
"data": {
|
@@ -922,8 +907,8 @@
|
|
922 |
"\n",
|
923 |
" <div>\n",
|
924 |
" \n",
|
925 |
-
" <progress value='
|
926 |
-
" [
|
927 |
" </div>\n",
|
928 |
" <table border=\"1\" class=\"dataframe\">\n",
|
929 |
" <thead>\n",
|
@@ -935,43 +920,43 @@
|
|
935 |
" <tbody>\n",
|
936 |
" <tr>\n",
|
937 |
" <td>100</td>\n",
|
938 |
-
" <td>1.
|
939 |
" </tr>\n",
|
940 |
" <tr>\n",
|
941 |
" <td>200</td>\n",
|
942 |
-
" <td>1.
|
943 |
" </tr>\n",
|
944 |
" <tr>\n",
|
945 |
" <td>300</td>\n",
|
946 |
-
" <td>1.
|
947 |
" </tr>\n",
|
948 |
" <tr>\n",
|
949 |
" <td>400</td>\n",
|
950 |
-
" <td>1.
|
951 |
" </tr>\n",
|
952 |
" <tr>\n",
|
953 |
" <td>500</td>\n",
|
954 |
-
" <td>1.
|
955 |
" </tr>\n",
|
956 |
" <tr>\n",
|
957 |
" <td>600</td>\n",
|
958 |
-
" <td>1.
|
959 |
" </tr>\n",
|
960 |
" <tr>\n",
|
961 |
" <td>700</td>\n",
|
962 |
-
" <td>1.
|
963 |
" </tr>\n",
|
964 |
" <tr>\n",
|
965 |
" <td>800</td>\n",
|
966 |
-
" <td>1.
|
967 |
" </tr>\n",
|
968 |
" <tr>\n",
|
969 |
" <td>900</td>\n",
|
970 |
-
" <td>1.
|
971 |
" </tr>\n",
|
972 |
" <tr>\n",
|
973 |
" <td>1000</td>\n",
|
974 |
-
" <td>1.
|
975 |
" </tr>\n",
|
976 |
" </tbody>\n",
|
977 |
"</table><p>"
|
@@ -996,10 +981,55 @@
|
|
996 |
},
|
997 |
{
|
998 |
"cell_type": "code",
|
999 |
-
"execution_count":
|
1000 |
"id": "263cc15e-8e9d-4bd8-9708-ec1638bc1165",
|
1001 |
"metadata": {},
|
1002 |
-
"outputs": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1003 |
"source": [
|
1004 |
"from peft import PeftModel\n",
|
1005 |
"from transformers import AutoModelForCausalLM\n",
|
@@ -1012,13 +1042,526 @@
|
|
1012 |
},
|
1013 |
{
|
1014 |
"cell_type": "code",
|
1015 |
-
"execution_count":
|
1016 |
"id": "8eef4f4f-52da-4ba9-8a22-2b7874420562",
|
1017 |
-
"metadata": {
|
1018 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1019 |
"source": [
|
1020 |
"model.push_to_hub('aissatoubalde/lab')"
|
1021 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1022 |
}
|
1023 |
],
|
1024 |
"metadata": {
|
|
|
547 |
},
|
548 |
{
|
549 |
"cell_type": "code",
|
550 |
+
"execution_count": 26,
|
551 |
"id": "18d5599f-992d-4d8e-a90c-4d43774be473",
|
552 |
"metadata": {},
|
553 |
"outputs": [
|
|
|
555 |
"name": "stdout",
|
556 |
"output_type": "stream",
|
557 |
"text": [
|
558 |
+
"trainable params: 18,350,080 || all params: 2,798,033,920 || trainable%: 0.6558204984162593\n"
|
559 |
]
|
560 |
}
|
561 |
],
|
|
|
563 |
"config = LoraConfig(\n",
|
564 |
" r=16,\n",
|
565 |
" lora_alpha=16,\n",
|
566 |
+
" #target_modules=[\"q_proj\",\"k_proj\",\"v_proj\"],\n",
|
567 |
" lora_dropout=0.05,\n",
|
568 |
" bias=\"none\",\n",
|
569 |
" task_type=\"CAUSAL_LM\"\n",
|
|
|
575 |
},
|
576 |
{
|
577 |
"cell_type": "code",
|
578 |
+
"execution_count": 27,
|
579 |
"id": "baeee903-3dce-48b2-93c3-7a697d8c6daf",
|
580 |
"metadata": {},
|
581 |
"outputs": [],
|
|
|
587 |
},
|
588 |
{
|
589 |
"cell_type": "code",
|
590 |
+
"execution_count": 28,
|
591 |
"id": "28a9b24a-a822-4fcb-96b3-d77b7ea30a5f",
|
592 |
"metadata": {},
|
593 |
"outputs": [
|
594 |
+
{
|
595 |
+
"name": "stderr",
|
596 |
+
"output_type": "stream",
|
597 |
+
"text": [
|
598 |
+
"huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
|
599 |
+
"To disable this warning, you can either:\n",
|
600 |
+
"\t- Avoid using `tokenizers` before the fork if possible\n",
|
601 |
+
"\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n"
|
602 |
+
]
|
603 |
+
},
|
604 |
{
|
605 |
"name": "stdout",
|
606 |
"output_type": "stream",
|
|
|
619 |
},
|
620 |
{
|
621 |
"cell_type": "code",
|
622 |
+
"execution_count": 29,
|
623 |
"id": "1ee7fd2a-38e4-4f23-a978-0bdeeda64d8b",
|
624 |
"metadata": {},
|
625 |
"outputs": [],
|
|
|
634 |
},
|
635 |
{
|
636 |
"cell_type": "code",
|
637 |
+
"execution_count": 30,
|
638 |
"id": "e84c29e2-843e-42c2-8c0f-324d392e671c",
|
639 |
"metadata": {},
|
640 |
"outputs": [
|
|
|
642 |
"name": "stderr",
|
643 |
"output_type": "stream",
|
644 |
"text": [
|
645 |
+
"Tokenizing data: 100%|ββββββββββ| 16412/16412 [00:02<00:00, 7265.30 examples/s]\n"
|
646 |
]
|
647 |
},
|
648 |
{
|
|
|
654 |
"})"
|
655 |
]
|
656 |
},
|
657 |
+
"execution_count": 30,
|
658 |
"metadata": {},
|
659 |
"output_type": "execute_result"
|
660 |
}
|
|
|
669 |
},
|
670 |
{
|
671 |
"cell_type": "code",
|
672 |
+
"execution_count": 31,
|
673 |
"id": "ac968254-5338-49df-950d-222b82647407",
|
674 |
"metadata": {},
|
675 |
"outputs": [],
|
|
|
690 |
},
|
691 |
{
|
692 |
"cell_type": "code",
|
693 |
+
"execution_count": 32,
|
694 |
"id": "729df2d0-0890-4ac4-adf3-c167a6e9669d",
|
695 |
"metadata": {},
|
696 |
"outputs": [
|
|
|
729 |
},
|
730 |
{
|
731 |
"cell_type": "code",
|
732 |
+
"execution_count": 33,
|
733 |
"id": "525f81a4-eb92-466e-bb9c-cd63122231ab",
|
734 |
"metadata": {},
|
735 |
"outputs": [
|
|
|
747 |
"name": "stdout",
|
748 |
"output_type": "stream",
|
749 |
"text": [
|
750 |
+
"Requirement already satisfied: jupyterlab in /usr/local/lib/python3.11/site-packages (4.1.1)\n",
|
751 |
+
"Requirement already satisfied: jupyterlab-git in /usr/local/lib/python3.11/site-packages (0.50.0)\n",
|
|
|
|
|
|
|
752 |
"Requirement already satisfied: async-lru>=1.0.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (2.0.4)\n",
|
753 |
"Requirement already satisfied: httpx>=0.25.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (0.26.0)\n",
|
754 |
"Requirement already satisfied: ipykernel in /usr/local/lib/python3.11/site-packages (from jupyterlab) (6.29.1)\n",
|
|
|
761 |
"Requirement already satisfied: packaging in /usr/local/lib/python3.11/site-packages (from jupyterlab) (23.2)\n",
|
762 |
"Requirement already satisfied: tornado>=6.2.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab) (6.4)\n",
|
763 |
"Requirement already satisfied: traitlets in /usr/local/lib/python3.11/site-packages (from jupyterlab) (5.14.1)\n",
|
764 |
+
"Requirement already satisfied: nbdime~=4.0.1 in /usr/local/lib/python3.11/site-packages (from jupyterlab-git) (4.0.1)\n",
|
|
|
765 |
"Requirement already satisfied: nbformat in /usr/local/lib/python3.11/site-packages (from jupyterlab-git) (5.9.2)\n",
|
766 |
"Requirement already satisfied: pexpect in /usr/local/lib/python3.11/site-packages (from jupyterlab-git) (4.9.0)\n",
|
767 |
"Requirement already satisfied: anyio in /usr/local/lib/python3.11/site-packages (from httpx>=0.25.0->jupyterlab) (4.2.0)\n",
|
|
|
787 |
"Requirement already satisfied: json5>=0.9.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (0.9.14)\n",
|
788 |
"Requirement already satisfied: jsonschema>=4.18.0 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (4.21.1)\n",
|
789 |
"Requirement already satisfied: requests>=2.31 in /usr/local/lib/python3.11/site-packages (from jupyterlab-server<3,>=2.19.0->jupyterlab) (2.31.0)\n",
|
790 |
+
"Requirement already satisfied: colorama in /usr/local/lib/python3.11/site-packages (from nbdime~=4.0.1->jupyterlab-git) (0.4.6)\n",
|
791 |
+
"Requirement already satisfied: gitpython!=2.1.4,!=2.1.5,!=2.1.6 in /usr/local/lib/python3.11/site-packages (from nbdime~=4.0.1->jupyterlab-git) (3.1.41)\n",
|
792 |
+
"Requirement already satisfied: jupyter-server-mathjax>=0.2.2 in /usr/local/lib/python3.11/site-packages (from nbdime~=4.0.1->jupyterlab-git) (0.2.6)\n",
|
793 |
+
"Requirement already satisfied: pygments in /usr/local/lib/python3.11/site-packages (from nbdime~=4.0.1->jupyterlab-git) (2.17.2)\n",
|
|
|
|
|
|
|
|
|
794 |
"Requirement already satisfied: fastjsonschema in /usr/local/lib/python3.11/site-packages (from nbformat->jupyterlab-git) (2.19.1)\n",
|
795 |
"Requirement already satisfied: comm>=0.1.1 in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (0.2.1)\n",
|
796 |
"Requirement already satisfied: debugpy>=1.6.5 in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (1.8.0)\n",
|
|
|
799 |
"Requirement already satisfied: nest-asyncio in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (1.6.0)\n",
|
800 |
"Requirement already satisfied: psutil in /usr/local/lib/python3.11/site-packages (from ipykernel->jupyterlab) (5.9.8)\n",
|
801 |
"Requirement already satisfied: ptyprocess>=0.5 in /usr/local/lib/python3.11/site-packages (from pexpect->jupyterlab-git) (0.7.0)\n",
|
802 |
+
"Requirement already satisfied: gitdb<5,>=4.0.1 in /usr/local/lib/python3.11/site-packages (from gitpython!=2.1.4,!=2.1.5,!=2.1.6->nbdime~=4.0.1->jupyterlab-git) (4.0.11)\n",
|
|
|
803 |
"Requirement already satisfied: decorator in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (5.1.1)\n",
|
804 |
"Requirement already satisfied: jedi>=0.16 in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (0.19.1)\n",
|
805 |
"Requirement already satisfied: prompt-toolkit<3.1.0,>=3.0.41 in /usr/local/lib/python3.11/site-packages (from ipython>=7.23.1->ipykernel->jupyterlab) (3.0.43)\n",
|
|
|
826 |
"Requirement already satisfied: argon2-cffi-bindings in /usr/local/lib/python3.11/site-packages (from argon2-cffi->jupyter-server<3,>=2.4.0->jupyterlab) (21.2.0)\n",
|
827 |
"Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.11/site-packages (from bleach!=5.0.0->nbconvert>=6.4.4->jupyter-server<3,>=2.4.0->jupyterlab) (1.16.0)\n",
|
828 |
"Requirement already satisfied: webencodings in /usr/local/lib/python3.11/site-packages (from bleach!=5.0.0->nbconvert>=6.4.4->jupyter-server<3,>=2.4.0->jupyterlab) (0.5.1)\n",
|
829 |
+
"Requirement already satisfied: smmap<6,>=3.0.1 in /usr/local/lib/python3.11/site-packages (from gitdb<5,>=4.0.1->gitpython!=2.1.4,!=2.1.5,!=2.1.6->nbdime~=4.0.1->jupyterlab-git) (5.0.1)\n",
|
|
|
830 |
"Requirement already satisfied: parso<0.9.0,>=0.8.3 in /usr/local/lib/python3.11/site-packages (from jedi>=0.16->ipython>=7.23.1->ipykernel->jupyterlab) (0.8.3)\n",
|
831 |
"Requirement already satisfied: fqdn in /usr/local/lib/python3.11/site-packages (from jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (1.5.1)\n",
|
832 |
"Requirement already satisfied: isoduration in /usr/local/lib/python3.11/site-packages (from jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (20.11.0)\n",
|
|
|
842 |
"Requirement already satisfied: pycparser in /usr/local/lib/python3.11/site-packages (from cffi>=1.0.1->argon2-cffi-bindings->argon2-cffi->jupyter-server<3,>=2.4.0->jupyterlab) (2.21)\n",
|
843 |
"Requirement already satisfied: arrow>=0.15.0 in /usr/local/lib/python3.11/site-packages (from isoduration->jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (1.3.0)\n",
|
844 |
"Requirement already satisfied: types-python-dateutil>=2.8.10 in /usr/local/lib/python3.11/site-packages (from arrow>=0.15.0->isoduration->jsonschema[format-nongpl]>=4.18.0->jupyter-events>=0.9.0->jupyter-server<3,>=2.4.0->jupyterlab) (2.8.19.20240106)\n",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
845 |
"Note: you may need to restart the kernel to use updated packages.\n"
|
846 |
]
|
847 |
}
|
|
|
897 |
"cell_type": "code",
|
898 |
"execution_count": null,
|
899 |
"id": "3bf553b6-b26c-49c3-9407-74c8d53a395e",
|
900 |
+
"metadata": {
|
901 |
+
"scrolled": true
|
902 |
+
},
|
903 |
"outputs": [
|
904 |
{
|
905 |
"data": {
|
|
|
907 |
"\n",
|
908 |
" <div>\n",
|
909 |
" \n",
|
910 |
+
" <progress value='1090' max='1100' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
|
911 |
+
" [1090/1100 12:44 < 00:07, 1.42 it/s, Epoch 0.27/1]\n",
|
912 |
" </div>\n",
|
913 |
" <table border=\"1\" class=\"dataframe\">\n",
|
914 |
" <thead>\n",
|
|
|
920 |
" <tbody>\n",
|
921 |
" <tr>\n",
|
922 |
" <td>100</td>\n",
|
923 |
+
" <td>1.317300</td>\n",
|
924 |
" </tr>\n",
|
925 |
" <tr>\n",
|
926 |
" <td>200</td>\n",
|
927 |
+
" <td>1.162700</td>\n",
|
928 |
" </tr>\n",
|
929 |
" <tr>\n",
|
930 |
" <td>300</td>\n",
|
931 |
+
" <td>1.160700</td>\n",
|
932 |
" </tr>\n",
|
933 |
" <tr>\n",
|
934 |
" <td>400</td>\n",
|
935 |
+
" <td>1.096200</td>\n",
|
936 |
" </tr>\n",
|
937 |
" <tr>\n",
|
938 |
" <td>500</td>\n",
|
939 |
+
" <td>1.111000</td>\n",
|
940 |
" </tr>\n",
|
941 |
" <tr>\n",
|
942 |
" <td>600</td>\n",
|
943 |
+
" <td>1.088400</td>\n",
|
944 |
" </tr>\n",
|
945 |
" <tr>\n",
|
946 |
" <td>700</td>\n",
|
947 |
+
" <td>1.122100</td>\n",
|
948 |
" </tr>\n",
|
949 |
" <tr>\n",
|
950 |
" <td>800</td>\n",
|
951 |
+
" <td>1.102400</td>\n",
|
952 |
" </tr>\n",
|
953 |
" <tr>\n",
|
954 |
" <td>900</td>\n",
|
955 |
+
" <td>1.067200</td>\n",
|
956 |
" </tr>\n",
|
957 |
" <tr>\n",
|
958 |
" <td>1000</td>\n",
|
959 |
+
" <td>1.079000</td>\n",
|
960 |
" </tr>\n",
|
961 |
" </tbody>\n",
|
962 |
"</table><p>"
|
|
|
981 |
},
|
982 |
{
|
983 |
"cell_type": "code",
|
984 |
+
"execution_count": 24,
|
985 |
"id": "263cc15e-8e9d-4bd8-9708-ec1638bc1165",
|
986 |
"metadata": {},
|
987 |
+
"outputs": [
|
988 |
+
{
|
989 |
+
"name": "stderr",
|
990 |
+
"output_type": "stream",
|
991 |
+
"text": [
|
992 |
+
"Loading checkpoint shards: 100%|ββββββββββ| 2/2 [00:02<00:00, 1.34s/it]\n",
|
993 |
+
"adapter_config.json: 100%|ββββββββββ| 592/592 [00:00<00:00, 3.99MB/s]\n",
|
994 |
+
"adapter_model.safetensors: 100%|ββββββββββ| 31.5M/31.5M [00:00<00:00, 117MB/s]\n"
|
995 |
+
]
|
996 |
+
},
|
997 |
+
{
|
998 |
+
"data": {
|
999 |
+
"text/plain": [
|
1000 |
+
"PhiForCausalLM(\n",
|
1001 |
+
" (model): PhiModel(\n",
|
1002 |
+
" (embed_tokens): Embedding(51200, 2560)\n",
|
1003 |
+
" (embed_dropout): Dropout(p=0.0, inplace=False)\n",
|
1004 |
+
" (layers): ModuleList(\n",
|
1005 |
+
" (0-31): 32 x PhiDecoderLayer(\n",
|
1006 |
+
" (self_attn): PhiAttention(\n",
|
1007 |
+
" (q_proj): Linear(in_features=2560, out_features=2560, bias=True)\n",
|
1008 |
+
" (k_proj): Linear(in_features=2560, out_features=2560, bias=True)\n",
|
1009 |
+
" (v_proj): Linear(in_features=2560, out_features=2560, bias=True)\n",
|
1010 |
+
" (dense): Linear(in_features=2560, out_features=2560, bias=True)\n",
|
1011 |
+
" (rotary_emb): PhiRotaryEmbedding()\n",
|
1012 |
+
" )\n",
|
1013 |
+
" (mlp): PhiMLP(\n",
|
1014 |
+
" (activation_fn): NewGELUActivation()\n",
|
1015 |
+
" (fc1): Linear(in_features=2560, out_features=10240, bias=True)\n",
|
1016 |
+
" (fc2): Linear(in_features=10240, out_features=2560, bias=True)\n",
|
1017 |
+
" )\n",
|
1018 |
+
" (input_layernorm): LayerNorm((2560,), eps=1e-05, elementwise_affine=True)\n",
|
1019 |
+
" (resid_dropout): Dropout(p=0.1, inplace=False)\n",
|
1020 |
+
" )\n",
|
1021 |
+
" )\n",
|
1022 |
+
" (final_layernorm): LayerNorm((2560,), eps=1e-05, elementwise_affine=True)\n",
|
1023 |
+
" )\n",
|
1024 |
+
" (lm_head): Linear(in_features=2560, out_features=51200, bias=True)\n",
|
1025 |
+
")"
|
1026 |
+
]
|
1027 |
+
},
|
1028 |
+
"execution_count": 24,
|
1029 |
+
"metadata": {},
|
1030 |
+
"output_type": "execute_result"
|
1031 |
+
}
|
1032 |
+
],
|
1033 |
"source": [
|
1034 |
"from peft import PeftModel\n",
|
1035 |
"from transformers import AutoModelForCausalLM\n",
|
|
|
1042 |
},
|
1043 |
{
|
1044 |
"cell_type": "code",
|
1045 |
+
"execution_count": 25,
|
1046 |
"id": "8eef4f4f-52da-4ba9-8a22-2b7874420562",
|
1047 |
+
"metadata": {
|
1048 |
+
"scrolled": true
|
1049 |
+
},
|
1050 |
+
"outputs": [
|
1051 |
+
{
|
1052 |
+
"name": "stderr",
|
1053 |
+
"output_type": "stream",
|
1054 |
+
"text": [
|
1055 |
+
"model-00001-of-00003.safetensors: 0%| | 0.00/4.98G [00:00<?, ?B/s]\n",
|
1056 |
+
"model-00002-of-00003.safetensors: 0%| | 0.00/4.98G [00:00<?, ?B/s]\u001b[A\n",
|
1057 |
+
"\n",
|
1058 |
+
"model-00003-of-00003.safetensors: 0%| | 0.00/1.15G [00:00<?, ?B/s]\u001b[A\u001b[A\n",
|
1059 |
+
"\n",
|
1060 |
+
"\n",
|
1061 |
+
"model-00001-of-00003.safetensors: 0%| | 12.4M/4.98G [00:00<00:41, 121MB/s]\n",
|
1062 |
+
"\n",
|
1063 |
+
"model-00001-of-00003.safetensors: 0%| | 24.5M/4.98G [00:00<01:02, 79.2MB/s]\u001b[A\u001b[A\n",
|
1064 |
+
"model-00002-of-00003.safetensors: 0%| | 16.0M/4.98G [00:00<01:34, 52.4MB/s]\u001b[A\n",
|
1065 |
+
"\n",
|
1066 |
+
"model-00003-of-00003.safetensors: 1%|β | 16.0M/1.15G [00:00<00:26, 43.2MB/s]\u001b[A\u001b[A\n",
|
1067 |
+
"model-00001-of-00003.safetensors: 1%| | 33.2M/4.98G [00:00<01:30, 54.7MB/s]\u001b[A\n",
|
1068 |
+
"\n",
|
1069 |
+
"model-00003-of-00003.safetensors: 3%|β | 32.0M/1.15G [00:00<00:18, 61.3MB/s]\u001b[A\u001b[A\n",
|
1070 |
+
"model-00001-of-00003.safetensors: 1%| | 48.0M/4.98G [00:00<01:25, 57.8MB/s]\u001b[A\n",
|
1071 |
+
"\n",
|
1072 |
+
"model-00003-of-00003.safetensors: 4%|β | 48.0M/1.15G [00:00<00:18, 58.4MB/s]\u001b[A\u001b[A\n",
|
1073 |
+
"model-00001-of-00003.safetensors: 1%|β | 64.0M/4.98G [00:01<01:32, 53.0MB/s]\u001b[A\n",
|
1074 |
+
"model-00002-of-00003.safetensors: 2%|β | 80.0M/4.98G [00:01<01:14, 65.7MB/s]\u001b[A\n",
|
1075 |
+
"\n",
|
1076 |
+
"model-00001-of-00003.safetensors: 2%|β | 80.0M/4.98G [00:01<01:44, 47.0MB/s]\u001b[A\u001b[A\n",
|
1077 |
+
"model-00001-of-00003.safetensors: 2%|β | 96.0M/4.98G [00:01<01:31, 53.2MB/s]\u001b[A\n",
|
1078 |
+
"model-00001-of-00003.safetensors: 2%|β | 112M/4.98G [00:02<01:27, 55.7MB/s] \u001b[A\n",
|
1079 |
+
"model-00002-of-00003.safetensors: 3%|β | 128M/4.98G [00:02<01:21, 59.3MB/s]\u001b[A\n",
|
1080 |
+
"model-00001-of-00003.safetensors: 3%|β | 128M/4.98G [00:02<01:39, 48.8MB/s]\u001b[A\n",
|
1081 |
+
"model-00001-of-00003.safetensors: 3%|β | 160M/4.98G [00:02<01:19, 61.0MB/s]\u001b[A\n",
|
1082 |
+
"model-00002-of-00003.safetensors: 4%|β | 176M/4.98G [00:02<01:23, 57.6MB/s]\u001b[A\n",
|
1083 |
+
"model-00001-of-00003.safetensors: 4%|β | 192M/4.98G [00:03<01:16, 62.8MB/s]\u001b[A\n",
|
1084 |
+
"model-00002-of-00003.safetensors: 4%|β | 208M/4.98G [00:03<01:16, 62.6MB/s]\u001b[A\n",
|
1085 |
+
"model-00001-of-00003.safetensors: 4%|β | 224M/4.98G [00:03<01:19, 59.6MB/s]\u001b[A\n",
|
1086 |
+
"model-00001-of-00003.safetensors: 5%|β | 240M/4.98G [00:04<01:17, 61.4MB/s]\u001b[A\n",
|
1087 |
+
"model-00001-of-00003.safetensors: 5%|β | 256M/4.98G [00:04<01:17, 61.0MB/s]\u001b[A\n",
|
1088 |
+
"model-00001-of-00003.safetensors: 5%|β | 272M/4.98G [00:04<01:24, 55.7MB/s]\u001b[A\n",
|
1089 |
+
"\n",
|
1090 |
+
"model-00003-of-00003.safetensors: 7%|β | 80.0M/1.15G [00:04<01:41, 10.6MB/s]\u001b[A\u001b[A\n",
|
1091 |
+
"model-00001-of-00003.safetensors: 6%|β | 288M/4.98G [00:05<01:22, 57.2MB/s]\u001b[A\n",
|
1092 |
+
"\n",
|
1093 |
+
"model-00003-of-00003.safetensors: 8%|β | 96.0M/1.15G [00:05<01:13, 14.4MB/s]\u001b[A\u001b[A\n",
|
1094 |
+
"model-00001-of-00003.safetensors: 6%|β | 304M/4.98G [00:05<01:28, 53.1MB/s]\u001b[A\n",
|
1095 |
+
"model-00002-of-00003.safetensors: 6%|β | 313M/4.98G [00:05<01:24, 55.0MB/s]\u001b[A\n",
|
1096 |
+
"\n",
|
1097 |
+
"model-00001-of-00003.safetensors: 6%|β | 320M/4.98G [00:05<01:20, 58.1MB/s] \u001b[A\u001b[A\n",
|
1098 |
+
"model-00002-of-00003.safetensors: 6%|β | 320M/4.98G [00:05<01:39, 47.0MB/s]\u001b[A\n",
|
1099 |
+
"\n",
|
1100 |
+
"model-00001-of-00003.safetensors: 7%|β | 336M/4.98G [00:05<01:18, 59.1MB/s]\u001b[A\u001b[A\n",
|
1101 |
+
"model-00001-of-00003.safetensors: 7%|β | 352M/4.98G [00:06<01:17, 59.4MB/s]\u001b[A\n",
|
1102 |
+
"\n",
|
1103 |
+
"model-00003-of-00003.safetensors: 12%|ββ | 144M/1.15G [00:06<00:38, 26.2MB/s]\u001b[A\u001b[A\n",
|
1104 |
+
"model-00001-of-00003.safetensors: 8%|β | 384M/4.98G [00:06<01:07, 68.0MB/s]\u001b[A\n",
|
1105 |
+
"\n",
|
1106 |
+
"model-00003-of-00003.safetensors: 14%|ββ | 160M/1.15G [00:06<00:32, 31.0MB/s]\u001b[A\u001b[A\n",
|
1107 |
+
"model-00001-of-00003.safetensors: 8%|β | 400M/4.98G [00:06<01:11, 63.7MB/s]\u001b[A\n",
|
1108 |
+
"model-00002-of-00003.safetensors: 8%|β | 384M/4.98G [00:06<01:25, 53.8MB/s]\u001b[A\n",
|
1109 |
+
"\n",
|
1110 |
+
"model-00001-of-00003.safetensors: 8%|β | 416M/4.98G [00:07<01:11, 63.9MB/s]\u001b[A\u001b[A\n",
|
1111 |
+
"\n",
|
1112 |
+
"model-00003-of-00003.safetensors: 17%|ββ | 192M/1.15G [00:07<00:22, 42.4MB/s]\u001b[A\u001b[A\n",
|
1113 |
+
"model-00001-of-00003.safetensors: 9%|β | 432M/4.98G [00:07<01:10, 64.4MB/s]\u001b[A\n",
|
1114 |
+
"\n",
|
1115 |
+
"model-00003-of-00003.safetensors: 18%|ββ | 208M/1.15G [00:07<00:21, 45.0MB/s]\u001b[A\u001b[A\n",
|
1116 |
+
"model-00001-of-00003.safetensors: 9%|β | 448M/4.98G [00:07<01:09, 65.2MB/s]\u001b[A\n",
|
1117 |
+
"\n",
|
1118 |
+
"model-00003-of-00003.safetensors: 19%|ββ | 224M/1.15G [00:07<00:18, 49.2MB/s]\u001b[A\u001b[A\n",
|
1119 |
+
"\n",
|
1120 |
+
"model-00001-of-00003.safetensors: 9%|β | 464M/4.98G [00:07<01:20, 56.0MB/s]\u001b[A\u001b[A\n",
|
1121 |
+
"model-00002-of-00003.safetensors: 9%|β | 432M/4.98G [00:08<01:51, 41.0MB/s]\u001b[A\n",
|
1122 |
+
"model-00001-of-00003.safetensors: 10%|β | 480M/4.98G [00:08<01:18, 57.1MB/s]\u001b[A\n",
|
1123 |
+
"\n",
|
1124 |
+
"model-00003-of-00003.safetensors: 22%|βββ | 256M/1.15G [00:08<00:17, 52.1MB/s]\u001b[A\u001b[A\n",
|
1125 |
+
"model-00002-of-00003.safetensors: 9%|β | 441M/4.98G [00:08<01:50, 41.0MB/s]\u001b[A\n",
|
1126 |
+
"\n",
|
1127 |
+
"model-00001-of-00003.safetensors: 10%|β | 496M/4.98G [00:08<01:20, 56.0MB/s]\u001b[A\u001b[A\n",
|
1128 |
+
"model-00002-of-00003.safetensors: 9%|β | 448M/4.98G [00:08<02:36, 29.0MB/s]\u001b[A\n",
|
1129 |
+
"\n",
|
1130 |
+
"model-00001-of-00003.safetensors: 10%|β | 512M/4.98G [00:08<01:20, 55.2MB/s]\u001b[A\u001b[A\n",
|
1131 |
+
"model-00002-of-00003.safetensors: 9%|β | 464M/4.98G [00:08<01:58, 38.3MB/s]\u001b[A\n",
|
1132 |
+
"\n",
|
1133 |
+
"model-00001-of-00003.safetensors: 11%|β | 528M/4.98G [00:09<01:26, 51.3MB/s]\u001b[A\u001b[A\n",
|
1134 |
+
"model-00001-of-00003.safetensors: 11%|β | 544M/4.98G [00:09<01:22, 53.5MB/s]\u001b[A\n",
|
1135 |
+
"\n",
|
1136 |
+
"model-00003-of-00003.safetensors: 28%|βββ | 320M/1.15G [00:09<00:17, 46.6MB/s]\u001b[A\u001b[A\n",
|
1137 |
+
"model-00001-of-00003.safetensors: 11%|β | 560M/4.98G [00:09<01:16, 58.2MB/s]\u001b[A\n",
|
1138 |
+
"\n",
|
1139 |
+
"model-00003-of-00003.safetensors: 29%|βοΏ½οΏ½β | 336M/1.15G [00:09<00:17, 47.1MB/s]\u001b[A\u001b[A\n",
|
1140 |
+
"model-00001-of-00003.safetensors: 12%|ββ | 592M/4.98G [00:10<01:10, 62.1MB/s]\u001b[A\n",
|
1141 |
+
"\n",
|
1142 |
+
"model-00003-of-00003.safetensors: 31%|βββ | 352M/1.15G [00:10<00:16, 48.3MB/s]\u001b[A\u001b[A\n",
|
1143 |
+
"model-00002-of-00003.safetensors: 11%|β | 528M/4.98G [00:10<01:26, 51.2MB/s]\u001b[A\n",
|
1144 |
+
"model-00002-of-00003.safetensors: 11%|β | 544M/4.98G [00:10<01:19, 55.8MB/s]\u001b[A\n",
|
1145 |
+
"\n",
|
1146 |
+
"model-00001-of-00003.safetensors: 12%|ββ | 608M/4.98G [00:10<01:20, 54.3MB/s]\u001b[A\u001b[A\n",
|
1147 |
+
"model-00001-of-00003.safetensors: 13%|ββ | 624M/4.98G [00:10<01:13, 59.2MB/s]\u001b[A\n",
|
1148 |
+
"model-00002-of-00003.safetensors: 12%|ββ | 576M/4.98G [00:10<01:09, 63.7MB/s]\u001b[A\n",
|
1149 |
+
"\n",
|
1150 |
+
"model-00001-of-00003.safetensors: 13%|ββ | 640M/4.98G [00:11<01:15, 57.7MB/s]\u001b[A\u001b[A\n",
|
1151 |
+
"model-00002-of-00003.safetensors: 12%|ββ | 592M/4.98G [00:11<01:11, 61.1MB/s]\u001b[A\n",
|
1152 |
+
"\n",
|
1153 |
+
"model-00001-of-00003.safetensors: 13%|ββ | 656M/4.98G [00:11<01:17, 56.0MB/s]\u001b[A\u001b[A\n",
|
1154 |
+
"model-00002-of-00003.safetensors: 12%|ββ | 608M/4.98G [00:11<01:12, 60.3MB/s]\u001b[A\n",
|
1155 |
+
"\n",
|
1156 |
+
"model-00003-of-00003.safetensors: 36%|ββββ | 416M/1.15G [00:11<00:16, 45.4MB/s]\u001b[A\u001b[A\n",
|
1157 |
+
"model-00001-of-00003.safetensors: 13%|ββ | 672M/4.98G [00:11<01:19, 54.3MB/s]\u001b[A\n",
|
1158 |
+
"\n",
|
1159 |
+
"model-00003-of-00003.safetensors: 37%|ββββ | 432M/1.15G [00:11<00:14, 48.6MB/s]\u001b[A\u001b[A\n",
|
1160 |
+
"model-00002-of-00003.safetensors: 13%|ββ | 640M/4.98G [00:11<01:09, 62.3MB/s]\u001b[A\n",
|
1161 |
+
"\n",
|
1162 |
+
"model-00001-of-00003.safetensors: 14%|ββ | 688M/4.98G [00:12<01:36, 44.6MB/s]\u001b[A\u001b[A\n",
|
1163 |
+
"model-00002-of-00003.safetensors: 13%|ββ | 656M/4.98G [00:12<01:13, 59.0MB/s]\u001b[A\n",
|
1164 |
+
"\n",
|
1165 |
+
"model-00001-of-00003.safetensors: 14%|ββ | 704M/4.98G [00:12<01:28, 48.3MB/s]\u001b[A\u001b[A\n",
|
1166 |
+
"model-00001-of-00003.safetensors: 15%|ββ | 736M/4.98G [00:12<01:18, 53.9MB/s]\u001b[A\n",
|
1167 |
+
"model-00002-of-00003.safetensors: 14%|ββ | 688M/4.98G [00:12<01:30, 47.4MB/s]\u001b[A\n",
|
1168 |
+
"\n",
|
1169 |
+
"model-00001-of-00003.safetensors: 15%|ββ | 752M/4.98G [00:13<01:11, 58.9MB/s]\u001b[A\u001b[A\n",
|
1170 |
+
"model-00002-of-00003.safetensors: 14%|ββ | 704M/4.98G [00:13<01:26, 49.2MB/s]\u001b[A\n",
|
1171 |
+
"\n",
|
1172 |
+
"model-00003-of-00003.safetensors: 43%|βββββ | 496M/1.15G [00:13<00:14, 44.5MB/s]\u001b[A\u001b[A\n",
|
1173 |
+
"\n",
|
1174 |
+
"model-00003-of-00003.safetensors: 44%|βββββ | 512M/1.15G [00:13<00:13, 47.9MB/s]\u001b[A\u001b[A\n",
|
1175 |
+
"model-00001-of-00003.safetensors: 15%|ββ | 768M/4.98G [00:13<01:30, 46.6MB/s]\u001b[A\n",
|
1176 |
+
"\n",
|
1177 |
+
"model-00003-of-00003.safetensors: 46%|βββββ | 528M/1.15G [00:13<00:12, 48.7MB/s]\u001b[A\u001b[A\n",
|
1178 |
+
"model-00001-of-00003.safetensors: 16%|ββ | 784M/4.98G [00:13<01:25, 49.0MB/s]\u001b[A\n",
|
1179 |
+
"model-00002-of-00003.safetensors: 15%|ββ | 752M/4.98G [00:14<01:14, 56.8MB/s]\u001b[A\n",
|
1180 |
+
"\n",
|
1181 |
+
"model-00001-of-00003.safetensors: 16%|ββ | 816M/4.98G [00:14<01:15, 55.1MB/s]\u001b[A\u001b[A\n",
|
1182 |
+
"\n",
|
1183 |
+
"model-00003-of-00003.safetensors: 49%|βββββ | 560M/1.15G [00:14<00:12, 49.4MB/s]\u001b[A\u001b[A\n",
|
1184 |
+
"model-00002-of-00003.safetensors: 15%|ββ | 768M/4.98G [00:14<01:22, 51.2MB/s]\u001b[A\n",
|
1185 |
+
"\n",
|
1186 |
+
"model-00003-of-00003.safetensors: 50%|βββββ | 576M/1.15G [00:14<00:11, 51.9MB/s]\u001b[A\u001b[A\n",
|
1187 |
+
"\n",
|
1188 |
+
"model-00001-of-00003.safetensors: 17%|ββ | 848M/4.98G [00:15<01:27, 47.4MB/s]\u001b[A\u001b[A\n",
|
1189 |
+
"\n",
|
1190 |
+
"model-00001-of-00003.safetensors: 17%|ββ | 864M/4.98G [00:15<01:31, 45.0MB/s]\u001b[A\u001b[A\n",
|
1191 |
+
"\n",
|
1192 |
+
"model-00001-of-00003.safetensors: 18%|ββ | 880M/4.98G [00:15<01:22, 49.6MB/s]\u001b[A\u001b[A\n",
|
1193 |
+
"\n",
|
1194 |
+
"model-00001-of-00003.safetensors: 18%|ββ | 896M/4.98G [00:16<01:32, 44.3MB/s]\u001b[A\u001b[A\n",
|
1195 |
+
"\n",
|
1196 |
+
"model-00001-of-00003.safetensors: 18%|ββ | 912M/4.98G [00:16<01:24, 48.1MB/s]\u001b[A\u001b[A\n",
|
1197 |
+
"\n",
|
1198 |
+
"model-00001-of-00003.safetensors: 19%|ββ | 928M/4.98G [00:16<01:22, 49.4MB/s]\u001b[A\u001b[A\n",
|
1199 |
+
"\n",
|
1200 |
+
"model-00001-of-00003.safetensors: 19%|ββ | 944M/4.98G [00:17<01:20, 49.9MB/s]\u001b[A\u001b[A\n",
|
1201 |
+
"\n",
|
1202 |
+
"model-00001-of-00003.safetensors: 20%|ββ | 976M/4.98G [00:17<01:15, 53.2MB/s]\u001b[A\u001b[A\n",
|
1203 |
+
"\n",
|
1204 |
+
"model-00001-of-00003.safetensors: 20%|ββ | 992M/4.98G [00:18<01:13, 54.3MB/s]\u001b[A\u001b[A\n",
|
1205 |
+
"\n",
|
1206 |
+
"model-00001-of-00003.safetensors: 20%|ββ | 1.01G/4.98G [00:18<01:14, 53.6MB/s][A\u001b[A\n",
|
1207 |
+
"\n",
|
1208 |
+
"model-00001-of-00003.safetensors: 21%|ββ | 1.02G/4.98G [00:18<01:13, 54.1MB/s][A\u001b[A\n",
|
1209 |
+
"\n",
|
1210 |
+
"model-00001-of-00003.safetensors: 21%|ββ | 1.04G/4.98G [00:18<01:08, 57.6MB/s][A\u001b[A\n",
|
1211 |
+
"\n",
|
1212 |
+
"model-00001-of-00003.safetensors: 21%|ββ | 1.06G/4.98G [00:19<01:07, 57.8MB/s][A\u001b[A\n",
|
1213 |
+
"\n",
|
1214 |
+
"model-00001-of-00003.safetensors: 22%|βββ | 1.07G/4.98G [00:19<01:03, 61.4MB/s][A\u001b[A\n",
|
1215 |
+
"\n",
|
1216 |
+
"model-00003-of-00003.safetensors: 71%|βββββββ | 816M/1.15G [00:19<00:05, 58.2MB/s]\u001b[A\u001b[A\n",
|
1217 |
+
"model-00001-of-00003.safetensors: 22%|βββ | 1.09G/4.98G [00:19<01:03, 61.8MB/s][A\n",
|
1218 |
+
"\n",
|
1219 |
+
"model-00003-of-00003.safetensors: 72%|ββββββββ | 832M/1.15G [00:19<00:05, 59.7MB/s]\u001b[A\u001b[A\n",
|
1220 |
+
"model-00002-of-00003.safetensors: 16%|ββ | 800M/4.98G [00:19<05:43, 12.2MB/s]\u001b[A\n",
|
1221 |
+
"\n",
|
1222 |
+
"model-00001-of-00003.safetensors: 22%|βββ | 1.10G/4.98G [00:19<01:05, 59.0MB/s][A\u001b[A\n",
|
1223 |
+
"model-00002-of-00003.safetensors: 16%|ββ | 816M/4.98G [00:20<04:17, 16.2MB/s]\u001b[A\n",
|
1224 |
+
"\n",
|
1225 |
+
"model-00001-of-00003.safetensors: 22%|βββ | 1.12G/4.98G [00:20<01:07, 57.0MB/s][A\u001b[A\n",
|
1226 |
+
"model-00002-of-00003.safetensors: 17%|ββ | 832M/4.98G [00:20<03:19, 20.8MB/s]\u001b[A\n",
|
1227 |
+
"\n",
|
1228 |
+
"model-00003-of-00003.safetensors: 76%|ββββββββ | 880M/1.15G [00:20<00:04, 66.3MB/s]\u001b[A\u001b[A\n",
|
1229 |
+
"model-00001-of-00003.safetensors: 23%|βββ | 1.14G/4.98G [00:20<01:16, 50.4MB/s][A\n",
|
1230 |
+
"\n",
|
1231 |
+
"model-00001-of-00003.safetensors: 23%|βββ | 1.15G/4.98G [00:20<01:09, 55.0MB/s][A\u001b[A\n",
|
1232 |
+
"\n",
|
1233 |
+
"model-00001-of-00003.safetensors: 23%|βββ | 1.17G/4.98G [00:21<01:08, 55.6MB/s][A\u001b[A\n",
|
1234 |
+
"\n",
|
1235 |
+
"model-00003-of-00003.safetensors: 80%|ββββββββ | 928M/1.15G [00:21<00:03, 58.4MB/s]\u001b[A\u001b[A\n",
|
1236 |
+
"model-00001-of-00003.safetensors: 24%|βββ | 1.18G/4.98G [00:21<01:09, 54.9MB/s][A\n",
|
1237 |
+
"\n",
|
1238 |
+
"model-00001-of-00003.safetensors: 24%|βββ | 1.20G/4.98G [00:21<01:04, 59.0MB/s][A\u001b[A\n",
|
1239 |
+
"model-00002-of-00003.safetensors: 18%|ββ | 880M/4.98G [00:21<02:24, 28.4MB/s]\u001b[A\n",
|
1240 |
+
"\n",
|
1241 |
+
"model-00003-of-00003.safetensors: 83%|βββββββββ | 960M/1.15G [00:21<00:03, 57.5MB/s]\u001b[A\u001b[A\n",
|
1242 |
+
"model-00001-of-00003.safetensors: 24%|βββ | 1.22G/4.98G [00:22<01:09, 53.9MB/s][A\n",
|
1243 |
+
"\n",
|
1244 |
+
"model-00003-of-00003.safetensors: 85%|βββββββββ | 976M/1.15G [00:22<00:03, 56.4MB/s]\u001b[A\u001b[A\n",
|
1245 |
+
"model-00001-of-00003.safetensors: 25%|βββ | 1.23G/4.98G [00:22<01:05, 57.2MB/s][A\n",
|
1246 |
+
"\n",
|
1247 |
+
"model-00001-of-00003.safetensors: 25%|βββ | 1.25G/4.98G [00:22<00:59, 62.3MB/s][A\u001b[A\n",
|
1248 |
+
"model-00002-of-00003.safetensors: 19%|ββ | 928M/4.98G [00:22<01:32, 43.7MB/s]\u001b[A\n",
|
1249 |
+
"\n",
|
1250 |
+
"model-00001-of-00003.safetensors: 25%|βββ | 1.26G/4.98G [00:22<00:57, 64.8MB/s]\u001b[A\u001b[A\n",
|
1251 |
+
"model-00002-of-00003.safetensors: 19%|ββ | 944M/4.98G [00:22<01:25, 47.4MB/s]\u001b[A\n",
|
1252 |
+
"model-00002-of-00003.safetensors: 19%|ββ | 960M/4.98G [00:23<01:14, 53.7MB/s]\u001b[A\n",
|
1253 |
+
"\n",
|
1254 |
+
"model-00001-of-00003.safetensors: 26%|βββ | 1.28G/4.98G [00:23<01:05, 56.6MB/s]\u001b[A\u001b[A\n",
|
1255 |
+
"model-00002-of-00003.safetensors: 20%|ββ | 976M/4.98G [00:23<01:10, 56.7MB/s]\u001b[A\n",
|
1256 |
+
"\n",
|
1257 |
+
"model-00001-of-00003.safetensors: 26%|βββ | 1.30G/4.98G [00:23<01:04, 56.8MB/s]\u001b[A\u001b[A\n",
|
1258 |
+
"\n",
|
1259 |
+
"model-00003-of-00003.safetensors: 92%|ββββββββββ| 1.06G/1.15G [00:23<00:01, 57.7MB/s]\u001b[A\u001b[A\n",
|
1260 |
+
"model-00001-of-00003.safetensors: 27%|βββ | 1.33G/4.98G [00:23<01:00, 60.6MB/s][A\n",
|
1261 |
+
"model-00002-of-00003.safetensors: 20%|ββ | 1.01G/4.98G [00:23<01:16, 52.0MB/s]\u001b[A\n",
|
1262 |
+
"\n",
|
1263 |
+
"model-00003-of-00003.safetensors: 93%|ββββββββββ| 1.07G/1.15G [00:23<00:01, 49.7MB/s]\u001b[A\u001b[A\n",
|
1264 |
+
"model-00002-of-00003.safetensors: 21%|ββ | 1.02G/4.98G [00:24<01:11, 55.7MB/s]\u001b[A\n",
|
1265 |
+
"\n",
|
1266 |
+
"model-00001-of-00003.safetensors: 27%|βββ | 1.34G/4.98G [00:24<01:03, 56.9MB/s]\u001b[A\u001b[A\n",
|
1267 |
+
"\n",
|
1268 |
+
"model-00001-of-00003.safetensors: 27%|βββ | 1.36G/4.98G [00:24<01:00, 59.9MB/s]\u001b[A\u001b[A\n",
|
1269 |
+
"model-00001-of-00003.safetensors: 28%|βββ | 1.38G/4.98G [00:24<01:01, 59.0MB/s]\u001b[A\n",
|
1270 |
+
"\n",
|
1271 |
+
"model-00003-of-00003.safetensors: 97%|ββββββββββ| 1.12G/1.15G [00:24<00:00, 53.2MB/s]\u001b[A\u001b[A\n",
|
1272 |
+
"model-00001-of-00003.safetensors: 28%|βββ | 1.39G/4.98G [00:25<01:04, 56.0MB/s]\u001b[A\n",
|
1273 |
+
"\n",
|
1274 |
+
"model-00003-of-00003.safetensors: 98%|ββββββββββ| 1.14G/1.15G [00:25<00:00, 55.6MB/s]\u001b[A\u001b[A\n",
|
1275 |
+
"model-00001-of-00003.safetensors: 28%|βββ | 1.41G/4.98G [00:25<01:06, 54.1MB/s]\u001b[A\n",
|
1276 |
+
"model-00002-of-00003.safetensors: 22%|βββ | 1.09G/4.98G [00:25<01:17, 50.4MB/s]\u001b[A\n",
|
1277 |
+
"\n",
|
1278 |
+
"model-00003-of-00003.safetensors: 100%|ββββββββββ| 1.15G/1.15G [00:25<00:00, 45.0MB/s]\u001b[A\u001b[A\n",
|
1279 |
+
"model-00001-of-00003.safetensors: 29%|βββ | 1.42G/4.98G [00:25<01:08, 52.1MB/s]\n",
|
1280 |
+
"model-00001-of-00003.safetensors: 29%|βββ | 1.44G/4.98G [00:25<01:03, 55.7MB/s]\u001b[A\n",
|
1281 |
+
"model-00001-of-00003.safetensors: 29%|βββ | 1.46G/4.98G [00:26<01:01, 57.4MB/s]\u001b[A\n",
|
1282 |
+
"model-00002-of-00003.safetensors: 23%|βββ | 1.14G/4.98G [00:26<01:18, 49.0MB/s]\u001b[A\n",
|
1283 |
+
"model-00001-of-00003.safetensors: 30%|βββ | 1.47G/4.98G [00:26<01:09, 50.8MB/s]\u001b[A\n",
|
1284 |
+
"model-00001-of-00003.safetensors: 30%|βββ | 1.49G/4.98G [00:26<01:00, 57.3MB/s]\u001b[A\n",
|
1285 |
+
"model-00001-of-00003.safetensors: 30%|βββ | 1.50G/4.98G [00:27<01:00, 57.5MB/s]\u001b[A\n",
|
1286 |
+
"model-00001-of-00003.safetensors: 31%|βββ | 1.54G/4.98G [00:27<01:01, 56.4MB/s]\u001b[A\n",
|
1287 |
+
"model-00002-of-00003.safetensors: 24%|βββ | 1.18G/4.98G [00:27<02:03, 30.8MB/s]\u001b[A\n",
|
1288 |
+
"model-00001-of-00003.safetensors: 31%|βββ | 1.55G/4.98G [00:27<00:59, 57.2MB/s]\u001b[A\n",
|
1289 |
+
"model-00001-of-00003.safetensors: 31%|ββββ | 1.57G/4.98G [00:28<00:56, 60.5MB/s]\u001b[A\n",
|
1290 |
+
"model-00001-of-00003.safetensors: 32%|ββββ | 1.58G/4.98G [00:28<00:53, 63.7MB/s]\u001b[A\n",
|
1291 |
+
"model-00002-of-00003.safetensors: 24%|βββ | 1.22G/4.98G [00:28<01:49, 34.3MB/s]\u001b[A\n",
|
1292 |
+
"model-00001-of-00003.safetensors: 32%|ββββ | 1.62G/4.98G [00:28<00:52, 63.5MB/s]\u001b[A\n",
|
1293 |
+
"model-00002-of-00003.safetensors: 25%|βββ | 1.23G/4.98G [00:28<01:37, 38.6MB/s]\u001b[A\n",
|
1294 |
+
"model-00001-of-00003.safetensors: 33%|ββββ | 1.63G/4.98G [00:29<00:54, 61.5MB/s]\u001b[A\n",
|
1295 |
+
"model-00002-of-00003.safetensors: 25%|βββ | 1.25G/4.98G [00:29<01:39, 37.6MB/s]\u001b[A\n",
|
1296 |
+
"model-00001-of-00003.safetensors: 33%|ββββ | 1.65G/4.98G [00:29<00:53, 62.1MB/s]\u001b[A\n",
|
1297 |
+
"model-00002-of-00003.safetensors: 25%|βββ | 1.26G/4.98G [00:29<01:28, 42.1MB/s]\u001b[A\n",
|
1298 |
+
"model-00001-of-00003.safetensors: 33%|ββββ | 1.66G/4.98G [00:29<00:58, 56.8MB/s]\u001b[A\n",
|
1299 |
+
"model-00002-of-00003.safetensors: 26%|βββ | 1.29G/4.98G [00:29<01:21, 45.5MB/s]\u001b[A\n",
|
1300 |
+
"model-00002-of-00003.safetensors: 26%|βββ | 1.30G/4.98G [00:30<01:10, 52.1MB/s]\u001b[A\n",
|
1301 |
+
"model-00001-of-00003.safetensors: 34%|ββββ | 1.68G/4.98G [00:30<01:20, 41.3MB/s]\u001b[A\n",
|
1302 |
+
"model-00001-of-00003.safetensors: 37%|ββββ | 1.82G/4.98G [00:32<00:52, 60.2MB/s]\u001b[A\n",
|
1303 |
+
"model-00002-of-00003.safetensors: 27%|βββ | 1.33G/4.98G [00:32<04:56, 12.3MB/s]\u001b[A\n",
|
1304 |
+
"model-00001-of-00003.safetensors: 37%|ββββ | 1.84G/4.98G [00:33<00:52, 59.5MB/s]\u001b[A\n",
|
1305 |
+
"model-00001-of-00003.safetensors: 37%|ββββ | 1.86G/4.98G [00:33<00:55, 56.6MB/s]\u001b[A\n",
|
1306 |
+
"model-00002-of-00003.safetensors: 27%|βββ | 1.36G/4.98G [00:33<02:33, 23.6MB/s]\u001b[A\n",
|
1307 |
+
"model-00001-of-00003.safetensors: 38%|ββββ | 1.87G/4.98G [00:33<01:00, 51.6MB/s]\u001b[A\n",
|
1308 |
+
"model-00001-of-00003.safetensors: 38%|ββββ | 1.89G/4.98G [00:34<01:05, 47.0MB/s]\u001b[A\n",
|
1309 |
+
"model-00001-of-00003.safetensors: 38%|ββββ | 1.90G/4.98G [00:34<01:02, 49.6MB/s]\u001b[A\n",
|
1310 |
+
"model-00001-of-00003.safetensors: 39%|ββββ | 1.92G/4.98G [00:34<00:57, 52.9MB/s]\u001b[A\n",
|
1311 |
+
"model-00001-of-00003.safetensors: 39%|ββββ | 1.95G/4.98G [00:35<00:56, 53.8MB/s]\u001b[A\n",
|
1312 |
+
"model-00001-of-00003.safetensors: 40%|ββββ | 1.98G/4.98G [00:35<00:49, 60.1MB/s]\u001b[A\n",
|
1313 |
+
"model-00001-of-00003.safetensors: 40%|ββββ | 2.00G/4.98G [00:36<00:52, 56.5MB/s]\u001b[A\n",
|
1314 |
+
"model-00001-of-00003.safetensors: 40%|ββββ | 2.02G/4.98G [00:36<00:52, 56.2MB/s]\u001b[A\n",
|
1315 |
+
"model-00001-of-00003.safetensors: 41%|ββββ | 2.03G/4.98G [00:36<00:55, 53.6MB/s]\u001b[A\n",
|
1316 |
+
"model-00001-of-00003.safetensors: 41%|ββββ | 2.05G/4.98G [00:37<00:54, 53.7MB/s]\u001b[A\n",
|
1317 |
+
"model-00002-of-00003.safetensors: 31%|βββ | 1.54G/4.98G [00:37<01:23, 41.5MB/s]\u001b[A\n",
|
1318 |
+
"model-00001-of-00003.safetensors: 41%|βββββ | 2.06G/4.98G [00:37<01:11, 40.7MB/s]\u001b[A\n",
|
1319 |
+
"model-00001-of-00003.safetensors: 42%|βββββ | 2.08G/4.98G [00:38<01:03, 45.9MB/s]\u001b[A\n",
|
1320 |
+
"model-00001-of-00003.safetensors: 42%|βββββ | 2.10G/4.98G [00:38<00:57, 49.9MB/s]\u001b[A\n",
|
1321 |
+
"model-00001-of-00003.safetensors: 42%|βββββ | 2.11G/4.98G [00:38<00:58, 49.4MB/s]\u001b[A\n",
|
1322 |
+
"model-00001-of-00003.safetensors: 43%|βββββ | 2.13G/4.98G [00:38<00:53, 53.4MB/s]\u001b[A\n",
|
1323 |
+
"model-00001-of-00003.safetensors: 43%|βββββ | 2.14G/4.98G [00:39<00:55, 51.4MB/s]\u001b[A\n",
|
1324 |
+
"model-00001-of-00003.safetensors: 43%|βββββ | 2.16G/4.98G [00:39<00:54, 52.1MB/s]\u001b[A\n",
|
1325 |
+
"model-00001-of-00003.safetensors: 44%|βββββ | 2.18G/4.98G [00:39<00:49, 56.9MB/s]\u001b[A\n",
|
1326 |
+
"model-00001-of-00003.safetensors: 45%|βββββ | 2.22G/4.98G [00:40<00:51, 53.3MB/s]\u001b[A\n",
|
1327 |
+
"model-00001-of-00003.safetensors: 45%|βββββ | 2.26G/4.98G [00:41<00:46, 59.1MB/s]\u001b[A\n",
|
1328 |
+
"model-00001-of-00003.safetensors: 46%|βββββ | 2.27G/4.98G [00:41<00:43, 62.8MB/s]\u001b[A\n",
|
1329 |
+
"model-00001-of-00003.safetensors: 46%|βββββ | 2.29G/4.98G [00:41<00:44, 60.6MB/s]\u001b[A\n",
|
1330 |
+
"model-00002-of-00003.safetensors: 35%|ββββ | 1.74G/4.98G [00:41<01:10, 46.2MB/s]\u001b[A\n",
|
1331 |
+
"model-00001-of-00003.safetensors: 46%|βββββ | 2.30G/4.98G [00:41<00:45, 58.6MB/s]\u001b[A\n",
|
1332 |
+
"model-00001-of-00003.safetensors: 47%|βββββ | 2.34G/4.98G [00:42<00:44, 59.5MB/s]\u001b[A\n",
|
1333 |
+
"model-00001-of-00003.safetensors: 47%|βββββ | 2.35G/4.98G [00:42<00:45, 57.2MB/s]\u001b[A\n",
|
1334 |
+
"model-00002-of-00003.safetensors: 36%|ββββ | 1.81G/4.98G [00:42<01:07, 47.0MB/s]\u001b[A\n",
|
1335 |
+
"model-00001-of-00003.safetensors: 48%|βββββ | 2.37G/4.98G [00:43<00:56, 46.5MB/s]\u001b[A\n",
|
1336 |
+
"model-00001-of-00003.safetensors: 48%|βββββ | 2.38G/4.98G [00:43<00:51, 50.2MB/s]\u001b[A\n",
|
1337 |
+
"model-00001-of-00003.safetensors: 48%|βββββ | 2.40G/4.98G [00:43<00:48, 53.2MB/s]\u001b[A\n",
|
1338 |
+
"model-00001-of-00003.safetensors: 49%|βββββ | 2.43G/4.98G [00:44<00:49, 51.5MB/s]\u001b[A\n",
|
1339 |
+
"model-00002-of-00003.safetensors: 38%|ββββ | 1.89G/4.98G [00:44<01:03, 49.1MB/s]\u001b[A\n",
|
1340 |
+
"model-00001-of-00003.safetensors: 49%|βββββ | 2.45G/4.98G [00:44<00:54, 46.4MB/s]\u001b[A\n",
|
1341 |
+
"model-00001-of-00003.safetensors: 50%|βββββ | 2.48G/4.98G [00:45<00:45, 55.0MB/s]\u001b[A\n",
|
1342 |
+
"model-00001-of-00003.safetensors: 50%|βββββ | 2.50G/4.98G [00:45<00:46, 53.4MB/s]\u001b[A\n",
|
1343 |
+
"model-00001-of-00003.safetensors: 50%|βββββ | 2.51G/4.98G [00:45<00:47, 52.0MB/s]\u001b[A\n",
|
1344 |
+
"model-00002-of-00003.safetensors: 39%|ββββ | 1.97G/4.98G [00:45<00:59, 50.8MB/s]\u001b[A\n",
|
1345 |
+
"model-00001-of-00003.safetensors: 51%|βββββ | 2.54G/4.98G [00:46<00:47, 51.8MB/s]\u001b[A\n",
|
1346 |
+
"model-00001-of-00003.safetensors: 52%|ββββββ | 2.58G/4.98G [00:47<00:48, 50.1MB/s]\u001b[A\n",
|
1347 |
+
"model-00001-of-00003.safetensors: 52%|ββββββ | 2.59G/4.98G [00:47<00:44, 53.4MB/s]\u001b[A\n",
|
1348 |
+
"model-00001-of-00003.safetensors: 53%|ββββββ | 2.62G/4.98G [00:47<00:41, 57.2MB/s]\u001b[A\n",
|
1349 |
+
"model-00001-of-00003.safetensors: 53%|ββββββ | 2.64G/4.98G [00:48<00:39, 59.0MB/s]\u001b[A\n",
|
1350 |
+
"model-00002-of-00003.safetensors: 41%|βββββ | 2.06G/4.98G [00:48<01:03, 46.3MB/s]\u001b[A\n",
|
1351 |
+
"model-00001-of-00003.safetensors: 54%|ββββββ | 2.69G/4.98G [00:49<00:39, 57.9MB/s]\u001b[A\n",
|
1352 |
+
"model-00001-of-00003.safetensors: 54%|ββββββ | 2.70G/4.98G [00:49<00:38, 58.9MB/s]\u001b[A\n",
|
1353 |
+
"model-00001-of-00003.safetensors: 55%|ββββββ | 2.74G/4.98G [00:49<00:34, 64.5MB/s]\u001b[A\n",
|
1354 |
+
"model-00001-of-00003.safetensors: 55%|ββββββ | 2.75G/4.98G [00:50<00:33, 66.7MB/s]\u001b[A\n",
|
1355 |
+
"model-00002-of-00003.safetensors: 43%|βββββ | 2.14G/4.98G [00:50<00:59, 47.9MB/s]\u001b[A\n",
|
1356 |
+
"model-00002-of-00003.safetensors: 43%|βββββ | 2.16G/4.98G [00:50<00:55, 51.2MB/s]\u001b[A\n",
|
1357 |
+
"model-00002-of-00003.safetensors: 44%|βββββ | 2.18G/4.98G [00:50<00:52, 53.6MB/s]\u001b[A\n",
|
1358 |
+
"model-00001-of-00003.safetensors: 56%|ββββββ | 2.77G/4.98G [00:51<01:17, 28.6MB/s]\u001b[A\n",
|
1359 |
+
"model-00001-of-00003.safetensors: 56%|ββββββ | 2.78G/4.98G [00:51<01:08, 31.9MB/s]\u001b[A\n",
|
1360 |
+
"model-00001-of-00003.safetensors: 57%|ββββββ | 2.82G/4.98G [00:52<00:50, 42.8MB/s]\u001b[A\n",
|
1361 |
+
"model-00001-of-00003.safetensors: 57%|ββββββ | 2.85G/4.98G [00:52<00:47, 44.7MB/s]\u001b[A\n",
|
1362 |
+
"model-00001-of-00003.safetensors: 57%|ββββββ | 2.86G/4.98G [00:53<00:43, 48.2MB/s]\u001b[A\n",
|
1363 |
+
"model-00002-of-00003.safetensors: 46%|βββββ | 2.27G/4.98G [00:53<01:18, 34.4MB/s]\u001b[A\n",
|
1364 |
+
"model-00001-of-00003.safetensors: 58%|ββββββ | 2.88G/4.98G [00:53<00:54, 38.3MB/s]\u001b[A\n",
|
1365 |
+
"model-00001-of-00003.safetensors: 58%|ββββββ | 2.90G/4.98G [00:53<00:45, 45.9MB/s]\u001b[A\n",
|
1366 |
+
"model-00001-of-00003.safetensors: 58%|ββββββ | 2.91G/4.98G [00:54<00:43, 47.2MB/s]\u001b[A\n",
|
1367 |
+
"model-00002-of-00003.safetensors: 47%|βββββ | 2.34G/4.98G [00:54<00:51, 51.3MB/s]\u001b[A\n",
|
1368 |
+
"model-00001-of-00003.safetensors: 59%|ββββββ | 2.93G/4.98G [00:54<00:45, 45.4MB/s]\u001b[A\n",
|
1369 |
+
"model-00001-of-00003.safetensors: 59%|ββββββ | 2.94G/4.98G [00:55<00:44, 46.2MB/s]\u001b[A\n",
|
1370 |
+
"model-00001-of-00003.safetensors: 59%|ββββββ | 2.96G/4.98G [00:55<00:40, 49.6MB/s]\u001b[A\n",
|
1371 |
+
"model-00001-of-00003.safetensors: 60%|ββββββ | 2.99G/4.98G [00:55<00:33, 59.5MB/s]\u001b[A\n",
|
1372 |
+
"model-00001-of-00003.safetensors: 60%|ββββββ | 3.01G/4.98G [00:55<00:32, 61.7MB/s]\u001b[A\n",
|
1373 |
+
"model-00001-of-00003.safetensors: 61%|ββββββ | 3.02G/4.98G [00:56<00:31, 61.9MB/s]\u001b[A\n",
|
1374 |
+
"model-00001-of-00003.safetensors: 61%|βββββββ | 3.06G/4.98G [00:56<00:31, 60.2MB/s]\u001b[A\n",
|
1375 |
+
"model-00001-of-00003.safetensors: 62%|βββββββ | 3.09G/4.98G [00:57<00:31, 59.6MB/s]\u001b[A\n",
|
1376 |
+
"model-00001-of-00003.safetensors: 62%|βββββββ | 3.10G/4.98G [00:57<00:30, 61.2MB/s]\u001b[A\n",
|
1377 |
+
"model-00001-of-00003.safetensors: 63%|βββββββ | 3.12G/4.98G [00:57<00:29, 62.1MB/s]\u001b[A\n",
|
1378 |
+
"model-00001-of-00003.safetensors: 63%|βββββββ | 3.14G/4.98G [00:58<00:28, 65.1MB/s]\u001b[A\n",
|
1379 |
+
"model-00001-of-00003.safetensors: 63%|βββββββ | 3.15G/4.98G [00:58<00:30, 60.4MB/s]\u001b[A\n",
|
1380 |
+
"model-00001-of-00003.safetensors: 64%|βββββββ | 3.17G/4.98G [00:58<00:30, 59.5MB/s]\u001b[A\n",
|
1381 |
+
"model-00001-of-00003.safetensors: 64%|βββββββ | 3.18G/4.98G [00:58<00:29, 60.8MB/s]\u001b[A\n",
|
1382 |
+
"model-00001-of-00003.safetensors: 65%|βββββββ | 3.22G/4.98G [00:59<00:33, 52.8MB/s]\u001b[A\n",
|
1383 |
+
"model-00001-of-00003.safetensors: 65%|βββββββ | 3.23G/4.98G [00:59<00:31, 55.1MB/s]\u001b[A\n",
|
1384 |
+
"model-00001-of-00003.safetensors: 65%|βββββββ | 3.25G/4.98G [01:00<00:30, 57.6MB/s]\u001b[A\n",
|
1385 |
+
"model-00001-of-00003.safetensors: 66%|βββββββ | 3.28G/4.98G [01:00<00:28, 60.7MB/s]\u001b[A\n",
|
1386 |
+
"model-00001-of-00003.safetensors: 66%|βββββββ | 3.30G/4.98G [01:00<00:27, 61.3MB/s]\u001b[A\n",
|
1387 |
+
"model-00002-of-00003.safetensors: 53%|ββββββ | 2.66G/4.98G [01:00<00:46, 49.6MB/s]\u001b[A\n",
|
1388 |
+
"model-00001-of-00003.safetensors: 66%|βββββββ | 3.31G/4.98G [01:01<00:28, 57.8MB/s]\u001b[A\n",
|
1389 |
+
"model-00001-of-00003.safetensors: 67%|βββββββ | 3.33G/4.98G [01:01<00:28, 57.3MB/s]\u001b[A\n",
|
1390 |
+
"model-00001-of-00003.safetensors: 68%|βββββββ | 3.38G/4.98G [01:02<00:27, 59.0MB/s]\u001b[A\n",
|
1391 |
+
"model-00001-of-00003.safetensors: 68%|βββββββ | 3.39G/4.98G [01:02<00:27, 58.6MB/s]\u001b[A\n",
|
1392 |
+
"model-00001-of-00003.safetensors: 68%|βββββββ | 3.41G/4.98G [01:02<00:28, 55.5MB/s]\u001b[A\n",
|
1393 |
+
"model-00001-of-00003.safetensors: 69%|βββββββ | 3.46G/4.98G [01:03<00:24, 63.3MB/s]\u001b[A\n",
|
1394 |
+
"model-00001-of-00003.safetensors: 70%|βββββββ | 3.47G/4.98G [01:03<00:23, 64.2MB/s]\u001b[A\n",
|
1395 |
+
"model-00001-of-00003.safetensors: 70%|βββββββ | 3.49G/4.98G [01:04<00:24, 62.1MB/s]\u001b[A\n",
|
1396 |
+
"model-00001-of-00003.safetensors: 70%|βββββββ | 3.50G/4.98G [01:04<00:27, 54.2MB/s]\u001b[A\n",
|
1397 |
+
"model-00001-of-00003.safetensors: 71%|βββββββ | 3.52G/4.98G [01:04<00:26, 56.2MB/s]\u001b[A\n",
|
1398 |
+
"model-00002-of-00003.safetensors: 57%|ββββββ | 2.83G/4.98G [01:04<00:48, 44.3MB/s]\u001b[A\n",
|
1399 |
+
"model-00001-of-00003.safetensors: 71%|ββββββββ | 3.55G/4.98G [01:05<00:32, 43.8MB/s]\u001b[A\n",
|
1400 |
+
"model-00001-of-00003.safetensors: 72%|ββββββββ | 3.57G/4.98G [01:05<00:29, 47.6MB/s]\u001b[A\n",
|
1401 |
+
"model-00001-of-00003.safetensors: 72%|ββββββββ | 3.58G/4.98G [01:06<00:26, 52.6MB/s]\u001b[A\n",
|
1402 |
+
"model-00001-of-00003.safetensors: 72%|ββββββββ | 3.60G/4.98G [01:06<00:25, 53.5MB/s]\u001b[A\n",
|
1403 |
+
"model-00001-of-00003.safetensors: 73%|ββββββββ | 3.62G/4.98G [01:06<00:24, 56.2MB/s]\u001b[A\n",
|
1404 |
+
"model-00002-of-00003.safetensors: 59%|ββββββ | 2.93G/4.98G [01:06<00:35, 57.9MB/s]\u001b[A\n",
|
1405 |
+
"model-00001-of-00003.safetensors: 73%|ββββββββ | 3.63G/4.98G [01:06<00:24, 56.1MB/s]\u001b[A\n",
|
1406 |
+
"model-00001-of-00003.safetensors: 73%|ββββββββ | 3.65G/4.98G [01:07<00:23, 56.7MB/s]\u001b[A\n",
|
1407 |
+
"model-00001-of-00003.safetensors: 74%|ββββββββ | 3.68G/4.98G [01:07<00:21, 61.0MB/s]\u001b[A\n",
|
1408 |
+
"model-00001-of-00003.safetensors: 74%|ββββββββ | 3.70G/4.98G [01:07<00:21, 60.1MB/s]\u001b[A\n",
|
1409 |
+
"model-00001-of-00003.safetensors: 75%|ββββββββ | 3.71G/4.98G [01:08<00:19, 64.1MB/s]\u001b[A\n",
|
1410 |
+
"model-00002-of-00003.safetensors: 61%|ββββββ | 3.02G/4.98G [01:08<00:33, 57.9MB/s]\u001b[A\n",
|
1411 |
+
"model-00001-of-00003.safetensors: 75%|ββββββββ | 3.74G/4.98G [01:08<00:20, 60.8MB/s]\u001b[A\n",
|
1412 |
+
"model-00001-of-00003.safetensors: 75%|ββββββββ | 3.76G/4.98G [01:09<00:20, 61.1MB/s]\u001b[A\n",
|
1413 |
+
"model-00001-of-00003.safetensors: 76%|ββββββββ | 3.79G/4.98G [01:09<00:17, 68.0MB/s]\u001b[A\n",
|
1414 |
+
"model-00001-of-00003.safetensors: 76%|ββββββββ | 3.81G/4.98G [01:09<00:16, 69.2MB/s]\u001b[A\n",
|
1415 |
+
"model-00001-of-00003.safetensors: 77%|ββββββββ | 3.82G/4.98G [01:09<00:17, 67.6MB/s]\u001b[A\n",
|
1416 |
+
"model-00001-of-00003.safetensors: 77%|ββββββββ | 3.86G/4.98G [01:10<00:16, 70.2MB/s]\u001b[A\n",
|
1417 |
+
"model-00001-of-00003.safetensors: 78%|ββββββββ | 3.87G/4.98G [01:10<00:21, 50.8MB/s]\u001b[A\n",
|
1418 |
+
"model-00001-of-00003.safetensors: 78%|ββββββββ | 3.89G/4.98G [01:11<00:19, 57.0MB/s]\u001b[A\n",
|
1419 |
+
"model-00001-of-00003.safetensors: 78%|ββββββββ | 3.90G/4.98G [01:11<00:19, 54.5MB/s]\u001b[A\n",
|
1420 |
+
"model-00001-of-00003.safetensors: 79%|ββββββββ | 3.92G/4.98G [01:11<00:19, 54.8MB/s]\u001b[A\n",
|
1421 |
+
"model-00001-of-00003.safetensors: 80%|ββββββββ | 3.97G/4.98G [01:12<00:17, 58.5MB/s]\u001b[A\n",
|
1422 |
+
"model-00001-of-00003.safetensors: 80%|ββββββββ | 3.98G/4.98G [01:12<00:18, 54.4MB/s]\u001b[A\n",
|
1423 |
+
"model-00001-of-00003.safetensors: 80%|ββββββββ | 4.00G/4.98G [01:13<00:18, 54.3MB/s]\u001b[A\n",
|
1424 |
+
"model-00001-of-00003.safetensors: 81%|ββββββββ | 4.02G/4.98G [01:13<00:17, 56.1MB/s]\u001b[A\n",
|
1425 |
+
"model-00001-of-00003.safetensors: 81%|ββββββββ | 4.03G/4.98G [01:13<00:16, 57.5MB/s]\u001b[A\n",
|
1426 |
+
"model-00001-of-00003.safetensors: 81%|ββββββββ | 4.05G/4.98G [01:13<00:16, 57.3MB/s]\u001b[A\n",
|
1427 |
+
"model-00001-of-00003.safetensors: 82%|βββββββββ | 4.06G/4.98G [01:14<00:15, 60.2MB/s]\u001b[A\n",
|
1428 |
+
"model-00001-of-00003.safetensors: 82%|βββββββββ | 4.08G/4.98G [01:14<00:13, 65.1MB/s]\u001b[A\n",
|
1429 |
+
"model-00001-of-00003.safetensors: 82%|βββββββββ | 4.10G/4.98G [01:14<00:14, 61.9MB/s]\u001b[A\n",
|
1430 |
+
"model-00001-of-00003.safetensors: 83%|βββββββββ | 4.11G/4.98G [01:14<00:13, 62.2MB/s]\u001b[A\n",
|
1431 |
+
"model-00001-of-00003.safetensors: 83%|βββββββββ | 4.13G/4.98G [01:15<00:15, 56.3MB/s]\u001b[A\n",
|
1432 |
+
"model-00002-of-00003.safetensors: 68%|βββββββ | 3.38G/4.98G [01:15<00:25, 63.5MB/s]\u001b[A\n",
|
1433 |
+
"model-00001-of-00003.safetensors: 83%|βββββββββ | 4.14G/4.98G [01:15<00:16, 50.8MB/s]\u001b[A\n",
|
1434 |
+
"model-00001-of-00003.safetensors: 83%|βββββββββ | 4.16G/4.98G [01:15<00:14, 55.5MB/s]\u001b[A\n",
|
1435 |
+
"model-00002-of-00003.safetensors: 69%|βββββββ | 3.42G/4.98G [01:16<00:25, 61.3MB/s]\u001b[A\n",
|
1436 |
+
"model-00001-of-00003.safetensors: 84%|βββββββββ | 4.18G/4.98G [01:16<00:17, 46.8MB/s]\u001b[A\n",
|
1437 |
+
"model-00002-of-00003.safetensors: 69%|βββββββ | 3.46G/4.98G [01:16<00:27, 55.4MB/s]\u001b[A\n",
|
1438 |
+
"model-00002-of-00003.safetensors: 70%|βββββββ | 3.47G/4.98G [01:16<00:26, 56.1MB/s]\u001b[A\n",
|
1439 |
+
"model-00002-of-00003.safetensors: 70%|βββββββ | 3.49G/4.98G [01:17<00:26, 57.1MB/s]\u001b[A\n",
|
1440 |
+
"model-00002-of-00003.safetensors: 70%|βββββββ | 3.50G/4.98G [01:17<00:26, 55.5MB/s]\u001b[A\n",
|
1441 |
+
"model-00002-of-00003.safetensors: 71%|βββββββ | 3.52G/4.98G [01:17<00:31, 46.1MB/s]\u001b[A\n",
|
1442 |
+
"model-00001-of-00003.safetensors: 84%|βββββββββ | 4.21G/4.98G [01:18<00:32, 24.2MB/s]\u001b[A\n",
|
1443 |
+
"model-00001-of-00003.safetensors: 85%|βββββββββ | 4.22G/4.98G [01:18<00:25, 30.3MB/s]\u001b[A\n",
|
1444 |
+
"model-00002-of-00003.safetensors: 72%|ββββββββ | 3.57G/4.98G [01:18<00:27, 50.8MB/s]\u001b[A\n",
|
1445 |
+
"model-00002-of-00003.safetensors: 72%|ββββββββ | 3.58G/4.98G [01:19<00:24, 56.7MB/s]\u001b[A\n",
|
1446 |
+
"model-00001-of-00003.safetensors: 85%|βββββββββ | 4.24G/4.98G [01:19<00:26, 27.9MB/s]\u001b[A\n",
|
1447 |
+
"model-00001-of-00003.safetensors: 85%|βββββββββ | 4.26G/4.98G [01:19<00:23, 31.0MB/s]\u001b[A\n",
|
1448 |
+
"model-00001-of-00003.safetensors: 86%|βββββββββ | 4.27G/4.98G [01:20<00:19, 37.1MB/s]\u001b[A\n",
|
1449 |
+
"model-00001-of-00003.safetensors: 86%|βββββββββ | 4.29G/4.98G [01:20<00:16, 41.2MB/s]\u001b[A\n",
|
1450 |
+
"model-00001-of-00003.safetensors: 86%|βββββββββ | 4.30G/4.98G [01:20<00:15, 44.8MB/s]\u001b[A\n",
|
1451 |
+
"model-00001-of-00003.safetensors: 87%|βββββββββ | 4.32G/4.98G [01:20<00:13, 50.0MB/s]\u001b[A\n",
|
1452 |
+
"model-00001-of-00003.safetensors: 87%|βββββββββ | 4.34G/4.98G [01:21<00:12, 52.4MB/s]\u001b[A\n",
|
1453 |
+
"model-00002-of-00003.safetensors: 75%|ββββββββ | 3.71G/4.98G [01:21<00:20, 62.6MB/s]\u001b[A\n",
|
1454 |
+
"model-00001-of-00003.safetensors: 88%|βββββββββ | 4.37G/4.98G [01:21<00:10, 58.1MB/s]\u001b[A\n",
|
1455 |
+
"model-00002-of-00003.safetensors: 75%|ββββββββ | 3.74G/4.98G [01:21<00:20, 61.6MB/s]\u001b[A\n",
|
1456 |
+
"model-00001-of-00003.safetensors: 88%|βββββββββ | 4.40G/4.98G [01:22<00:10, 57.1MB/s]\u001b[A\n",
|
1457 |
+
"model-00001-of-00003.safetensors: 89%|βββββββββ | 4.42G/4.98G [01:22<00:09, 61.7MB/s]\u001b[A\n",
|
1458 |
+
"model-00001-of-00003.safetensors: 89%|βββββββββ | 4.43G/4.98G [01:22<00:08, 63.7MB/s]\u001b[A\n",
|
1459 |
+
"model-00001-of-00003.safetensors: 89%|βββββββββ | 4.45G/4.98G [01:22<00:08, 64.0MB/s]\u001b[A\n",
|
1460 |
+
"model-00001-of-00003.safetensors: 90%|βββββββββ | 4.46G/4.98G [01:23<00:08, 64.0MB/s]\u001b[A\n",
|
1461 |
+
"model-00001-of-00003.safetensors: 90%|βββββββββ | 4.50G/4.98G [01:23<00:07, 68.2MB/s]\u001b[A\n",
|
1462 |
+
"model-00002-of-00003.safetensors: 77%|ββββββββ | 3.86G/4.98G [01:23<00:18, 61.7MB/s]\u001b[A\n",
|
1463 |
+
"model-00002-of-00003.safetensors: 78%|ββββββββ | 3.87G/4.98G [01:23<00:17, 61.8MB/s]\u001b[A\n",
|
1464 |
+
"model-00001-of-00003.safetensors: 91%|βββββββββ | 4.51G/4.98G [01:24<00:11, 39.9MB/s]\u001b[A\n",
|
1465 |
+
"model-00001-of-00003.safetensors: 91%|βββββββββ | 4.53G/4.98G [01:24<00:09, 46.0MB/s]\u001b[A\n",
|
1466 |
+
"model-00001-of-00003.safetensors: 91%|βββββββββ | 4.54G/4.98G [01:24<00:08, 50.6MB/s]\u001b[A\n",
|
1467 |
+
"model-00001-of-00003.safetensors: 92%|ββββββββββ| 4.56G/4.98G [01:25<00:08, 52.7MB/s]\u001b[A\n",
|
1468 |
+
"model-00001-of-00003.safetensors: 92%|ββββββββββ| 4.59G/4.98G [01:25<00:06, 64.9MB/s]\u001b[A\n",
|
1469 |
+
"model-00002-of-00003.safetensors: 80%|ββββββββ | 3.97G/4.98G [01:25<00:15, 64.1MB/s]\u001b[A\n",
|
1470 |
+
"model-00001-of-00003.safetensors: 92%|ββββββββββ| 4.61G/4.98G [01:25<00:06, 58.5MB/s]\u001b[A\n",
|
1471 |
+
"model-00001-of-00003.safetensors: 93%|ββββββββββ| 4.66G/4.98G [01:26<00:05, 60.5MB/s]\u001b[A\n",
|
1472 |
+
"model-00001-of-00003.safetensors: 94%|ββββββββββ| 4.67G/4.98G [01:27<00:05, 55.7MB/s]\u001b[A\n",
|
1473 |
+
"model-00001-of-00003.safetensors: 94%|ββββββββββ| 4.69G/4.98G [01:27<00:05, 57.7MB/s]\u001b[A\n",
|
1474 |
+
"model-00001-of-00003.safetensors: 94%|ββββββββββ| 4.70G/4.98G [01:27<00:04, 63.4MB/s]\u001b[A\n",
|
1475 |
+
"model-00001-of-00003.safetensors: 95%|ββββββββββ| 4.72G/4.98G [01:27<00:04, 63.1MB/s]\u001b[A\n",
|
1476 |
+
"model-00001-of-00003.safetensors: 95%|ββββββββββ| 4.74G/4.98G [01:28<00:04, 58.4MB/s]\u001b[A\n",
|
1477 |
+
"model-00001-of-00003.safetensors: 95%|ββββββββββ| 4.75G/4.98G [01:28<00:03, 59.5MB/s]\u001b[A\n",
|
1478 |
+
"model-00001-of-00003.safetensors: 96%|ββββββββββ| 4.77G/4.98G [01:28<00:03, 57.9MB/s]\u001b[A\n",
|
1479 |
+
"model-00001-of-00003.safetensors: 96%|ββββββββββ| 4.78G/4.98G [01:29<00:04, 48.8MB/s]\u001b[A\n",
|
1480 |
+
"model-00001-of-00003.safetensors: 96%|ββββββββββ| 4.80G/4.98G [01:29<00:03, 53.6MB/s]\u001b[A\n",
|
1481 |
+
"model-00001-of-00003.safetensors: 97%|ββββββββββ| 4.82G/4.98G [01:29<00:02, 57.8MB/s]\u001b[A\n",
|
1482 |
+
"model-00001-of-00003.safetensors: 97%|ββββββββββ| 4.83G/4.98G [01:29<00:02, 59.4MB/s]\u001b[A\n",
|
1483 |
+
"model-00001-of-00003.safetensors: 98%|ββββββββββ| 4.86G/4.98G [01:30<00:02, 41.8MB/s]\u001b[A\n",
|
1484 |
+
"model-00001-of-00003.safetensors: 98%|ββββββββββ| 4.88G/4.98G [01:31<00:02, 47.7MB/s]\u001b[A\n",
|
1485 |
+
"model-00001-of-00003.safetensors: 98%|ββββββββββ| 4.90G/4.98G [01:31<00:01, 46.2MB/s]\u001b[A\n",
|
1486 |
+
"model-00001-of-00003.safetensors: 99%|ββββββββββ| 4.91G/4.98G [01:31<00:01, 50.4MB/s]\u001b[A\n",
|
1487 |
+
"model-00001-of-00003.safetensors: 99%|ββββββββββ| 4.93G/4.98G [01:31<00:00, 58.1MB/s]\u001b[A\n",
|
1488 |
+
"model-00001-of-00003.safetensors: 99%|ββββββββββ| 4.94G/4.98G [01:32<00:00, 58.9MB/s]\u001b[A\n",
|
1489 |
+
"model-00001-of-00003.safetensors: 100%|ββββββββββ| 4.96G/4.98G [01:32<00:00, 42.6MB/s]\u001b[A\n",
|
1490 |
+
"model-00001-of-00003.safetensors: 100%|ββββββββββ| 4.98G/4.98G [01:33<00:00, 45.2MB/s]\u001b[A\n",
|
1491 |
+
"model-00001-of-00003.safetensors: 100%|ββββββββββ| 4.98G/4.98G [01:33<00:00, 53.4MB/s]\u001b[A\n",
|
1492 |
+
"\n",
|
1493 |
+
"model-00002-of-00003.safetensors: 87%|βββββββββ | 4.34G/4.98G [01:33<00:13, 48.2MB/s]\u001b[A\n",
|
1494 |
+
"\n",
|
1495 |
+
"\n",
|
1496 |
+
"Upload 3 LFS files: 33%|ββββ | 1/3 [01:33<03:06, 93.49s/it]\u001b[A\u001b[A\u001b[A\n",
|
1497 |
+
"model-00002-of-00003.safetensors: 87%|βββββββββ | 4.35G/4.98G [01:33<00:12, 52.5MB/s]\u001b[A\n",
|
1498 |
+
"model-00002-of-00003.safetensors: 88%|βββββββββ | 4.37G/4.98G [01:33<00:11, 54.7MB/s]\u001b[A\n",
|
1499 |
+
"model-00002-of-00003.safetensors: 88%|βββββββββ | 4.38G/4.98G [01:34<00:11, 52.1MB/s]\u001b[A\n",
|
1500 |
+
"model-00002-of-00003.safetensors: 88%|βββββββββ | 4.40G/4.98G [01:34<00:11, 49.8MB/s]\u001b[A\n",
|
1501 |
+
"model-00002-of-00003.safetensors: 89%|βββββββββ | 4.42G/4.98G [01:35<00:13, 43.0MB/s]\u001b[A\n",
|
1502 |
+
"model-00002-of-00003.safetensors: 89%|βββββββββ | 4.43G/4.98G [01:35<00:13, 41.0MB/s]\u001b[A\n",
|
1503 |
+
"model-00002-of-00003.safetensors: 89%|βββββββββ | 4.45G/4.98G [01:35<00:11, 46.0MB/s]\u001b[A\n",
|
1504 |
+
"model-00002-of-00003.safetensors: 90%|βββββββββ | 4.46G/4.98G [01:36<00:10, 49.6MB/s]\u001b[A\n",
|
1505 |
+
"model-00002-of-00003.safetensors: 90%|βββββββββ | 4.48G/4.98G [01:36<00:08, 58.5MB/s]\u001b[A\n",
|
1506 |
+
"model-00002-of-00003.safetensors: 90%|βββββββββ | 4.50G/4.98G [01:36<00:08, 58.8MB/s]\u001b[A\n",
|
1507 |
+
"model-00002-of-00003.safetensors: 91%|βββββββββ | 4.51G/4.98G [01:36<00:07, 61.3MB/s]\u001b[A\n",
|
1508 |
+
"model-00002-of-00003.safetensors: 91%|βββββββββ | 4.53G/4.98G [01:37<00:07, 61.9MB/s]\u001b[A\n",
|
1509 |
+
"model-00002-of-00003.safetensors: 91%|βββββββββ | 4.54G/4.98G [01:37<00:06, 64.6MB/s]\u001b[A\n",
|
1510 |
+
"model-00002-of-00003.safetensors: 92%|ββββββββββ| 4.56G/4.98G [01:37<00:06, 61.9MB/s]\u001b[A\n",
|
1511 |
+
"model-00002-of-00003.safetensors: 92%|ββββββββββ| 4.58G/4.98G [01:37<00:07, 56.0MB/s]\u001b[A\n",
|
1512 |
+
"model-00002-of-00003.safetensors: 92%|ββββββββββ| 4.59G/4.98G [01:38<00:06, 58.9MB/s]\u001b[A\n",
|
1513 |
+
"model-00002-of-00003.safetensors: 92%|ββββββββββ| 4.61G/4.98G [01:38<00:06, 60.2MB/s]\u001b[A\n",
|
1514 |
+
"model-00002-of-00003.safetensors: 93%|ββββββββββ| 4.62G/4.98G [01:38<00:05, 62.7MB/s]\u001b[A\n",
|
1515 |
+
"model-00002-of-00003.safetensors: 93%|ββββββββββ| 4.64G/4.98G [01:38<00:05, 58.5MB/s]\u001b[A\n",
|
1516 |
+
"model-00002-of-00003.safetensors: 93%|ββββββββββ| 4.66G/4.98G [01:39<00:05, 59.8MB/s]\u001b[A\n",
|
1517 |
+
"model-00002-of-00003.safetensors: 94%|ββββββββββ| 4.67G/4.98G [01:39<00:05, 56.7MB/s]\u001b[A\n",
|
1518 |
+
"model-00002-of-00003.safetensors: 94%|ββββββββββ| 4.69G/4.98G [01:39<00:05, 53.8MB/s]\u001b[A\n",
|
1519 |
+
"model-00002-of-00003.safetensors: 94%|ββββββββββ| 4.70G/4.98G [01:40<00:05, 55.2MB/s]\u001b[A\n",
|
1520 |
+
"model-00002-of-00003.safetensors: 95%|ββββββββββ| 4.72G/4.98G [01:40<00:05, 51.0MB/s]\u001b[A\n",
|
1521 |
+
"model-00002-of-00003.safetensors: 95%|ββββββββββ| 4.74G/4.98G [01:40<00:05, 48.6MB/s]\u001b[A\n",
|
1522 |
+
"model-00002-of-00003.safetensors: 95%|ββββββββββ| 4.75G/4.98G [01:41<00:04, 50.4MB/s]\u001b[A\n",
|
1523 |
+
"model-00002-of-00003.safetensors: 96%|ββββββββββ| 4.77G/4.98G [01:41<00:03, 54.0MB/s]\u001b[A\n",
|
1524 |
+
"model-00002-of-00003.safetensors: 96%|ββββββββββ| 4.78G/4.98G [01:41<00:03, 58.2MB/s]\u001b[A\n",
|
1525 |
+
"model-00002-of-00003.safetensors: 96%|ββββββββββ| 4.80G/4.98G [01:41<00:03, 55.1MB/s]\u001b[A\n",
|
1526 |
+
"model-00002-of-00003.safetensors: 97%|ββββββββββ| 4.82G/4.98G [01:42<00:03, 52.0MB/s]\u001b[A\n",
|
1527 |
+
"model-00002-of-00003.safetensors: 97%|ββββββββββ| 4.83G/4.98G [01:42<00:02, 52.8MB/s]\u001b[A\n",
|
1528 |
+
"model-00002-of-00003.safetensors: 97%|ββββββββββ| 4.85G/4.98G [01:42<00:02, 55.3MB/s]\u001b[A\n",
|
1529 |
+
"model-00002-of-00003.safetensors: 98%|οΏ½οΏ½βββββββββ| 4.86G/4.98G [01:43<00:02, 57.7MB/s]\u001b[A\n",
|
1530 |
+
"model-00002-of-00003.safetensors: 98%|ββββββββββ| 4.88G/4.98G [01:43<00:01, 57.2MB/s]\u001b[A\n",
|
1531 |
+
"model-00002-of-00003.safetensors: 98%|ββββββββββ| 4.90G/4.98G [01:43<00:01, 55.4MB/s]\u001b[A\n",
|
1532 |
+
"model-00002-of-00003.safetensors: 99%|ββββββββββ| 4.91G/4.98G [01:43<00:01, 54.2MB/s]\u001b[A\n",
|
1533 |
+
"model-00002-of-00003.safetensors: 99%|ββββββββββ| 4.93G/4.98G [01:44<00:01, 52.4MB/s]\u001b[A\n",
|
1534 |
+
"model-00002-of-00003.safetensors: 99%|ββββββββββ| 4.94G/4.98G [01:44<00:00, 56.1MB/s]\u001b[A\n",
|
1535 |
+
"model-00002-of-00003.safetensors: 100%|ββββββββββ| 4.96G/4.98G [01:44<00:00, 57.7MB/s]\u001b[A\n",
|
1536 |
+
"model-00002-of-00003.safetensors: 100%|ββββββββββ| 4.98G/4.98G [01:45<00:00, 47.4MB/s]\u001b[A\n",
|
1537 |
+
"\n",
|
1538 |
+
"\n",
|
1539 |
+
"\n",
|
1540 |
+
"Upload 3 LFS files: 100%|ββββββββββ| 3/3 [01:45<00:00, 35.14s/it]\u001b[A\u001b[A\u001b[A\n"
|
1541 |
+
]
|
1542 |
+
},
|
1543 |
+
{
|
1544 |
+
"data": {
|
1545 |
+
"text/plain": [
|
1546 |
+
"CommitInfo(commit_url='https://huggingface.co/aissatoubalde/lab/commit/a31ac628a1effc894512237599c8b851981d1e31', commit_message='Upload PhiForCausalLM', commit_description='', oid='a31ac628a1effc894512237599c8b851981d1e31', pr_url=None, pr_revision=None, pr_num=None)"
|
1547 |
+
]
|
1548 |
+
},
|
1549 |
+
"execution_count": 25,
|
1550 |
+
"metadata": {},
|
1551 |
+
"output_type": "execute_result"
|
1552 |
+
}
|
1553 |
+
],
|
1554 |
"source": [
|
1555 |
"model.push_to_hub('aissatoubalde/lab')"
|
1556 |
]
|
1557 |
+
},
|
1558 |
+
{
|
1559 |
+
"cell_type": "code",
|
1560 |
+
"execution_count": null,
|
1561 |
+
"id": "fa43c8e1-945e-41ef-b2f6-e90db1f50140",
|
1562 |
+
"metadata": {},
|
1563 |
+
"outputs": [],
|
1564 |
+
"source": []
|
1565 |
}
|
1566 |
],
|
1567 |
"metadata": {
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4664
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b97f3e3a61f154dd18e7d0d40e91564daa7cfb393ad45e0eb1adf8bad6659cc5
|
3 |
size 4664
|