diff --git a/.github/workflows/checks.yml b/.github/workflows/checks.yml
index 6ba7c728e..76ba930cd 100644
--- a/.github/workflows/checks.yml
+++ b/.github/workflows/checks.yml
@@ -234,8 +234,8 @@ jobs:
# - "Grokking_Demo"
# - "Head_Detector_Demo"
# - "Interactive_Neuroscope"
- # - "LLaMA"
- # - "LLaMA2_GPU_Quantized" # Requires quantization libs + too slow for CI timeout
+ - "LLaMA"
+ - "LLaMA2_GPU_Quantized" # Requires quantization libs + too slow for CI timeout
- "Main_Demo"
# - "No_Position_Experiment"
- "Othello_GPT"
@@ -267,15 +267,22 @@ jobs:
- name: Install dependencies
run: |
uv lock --check
- uv sync
+ uv sync --group quantization
- name: Install pandoc
uses: awalsh128/cache-apt-pkgs-action@latest
with:
packages: pandoc
version: 1.0
+ - name: Authenticate HuggingFace
+ if: env.HF_TOKEN != ''
+ run: uv run python -c "import os; from huggingface_hub import login; login(token=os.environ['HF_TOKEN'])"
+ env:
+ HF_TOKEN: ${{ secrets.HF_TOKEN }}
- name: Check Notebook Output Consistency
# Note: currently only checks notebooks we have specifically setup for this
run: pytest --nbval-sanitize-with demos/doc_sanitize.cfg demos/${{ matrix.notebook }}.ipynb
+ env:
+ HF_TOKEN: ${{ secrets.HF_TOKEN }}
build-docs:
diff --git a/demos/Attribution_Patching_Demo.ipynb b/demos/Attribution_Patching_Demo.ipynb
index e7adb828b..1559b02b1 100644
--- a/demos/Attribution_Patching_Demo.ipynb
+++ b/demos/Attribution_Patching_Demo.ipynb
@@ -252,13 +252,6 @@
}
},
"outputs": [
- {
- "name": "stderr",
- "output_type": "stream",
- "text": [
- "Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.\n"
- ]
- },
{
"data": {
"application/vnd.jupyter.widget-view+json": {
@@ -3807,7 +3800,7 @@
"Top 0th token. Logit: 20.73 Prob: 95.80% Token: | Paris|\n",
"Top 1th token. Logit: 16.49 Prob: 1.39% Token: | E|\n",
"Top 2th token. Logit: 14.69 Prob: 0.23% Token: | the|\n",
- "Top 3th token. Logit: 14.58 Prob: 0.21% Token: | \u00c9|\n",
+ "Top 3th token. Logit: 14.58 Prob: 0.21% Token: | É|\n",
"Top 4th token. Logit: 14.44 Prob: 0.18% Token: | France|\n",
"Top 5th token. Logit: 14.36 Prob: 0.16% Token: | Mont|\n",
"Top 6th token. Logit: 13.77 Prob: 0.09% Token: | Le|\n",
@@ -4237,11 +4230,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d230304b88114f2a9b85f5a48f441ce6",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_ee535e2cfa694be1a7857b1867b8b608",
"tabbable": null,
"tooltip": null,
- "value": "\u2007456k/?\u2007[00:00<00:00,\u200712.7MB/s]"
+ "value": " 456k/? [00:00<00:00, 12.7MB/s]"
}
},
"020cf001eb7d496295a325cbc0ee8718": {
@@ -4278,7 +4271,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_af67816d50074ae498ef9b600b4175ed",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7ac7b11ef3e34bf1a12926c745e08707",
"tabbable": null,
"tooltip": null,
@@ -4301,11 +4294,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_7a32c6104cdb4eee8dadc248c129040c",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_8c0e7d4b46c14e2bb2167752820a9274",
"tabbable": null,
"tooltip": null,
- "value": "\u20071.36M/?\u2007[00:00<00:00,\u200717.1MB/s]"
+ "value": " 1.36M/? [00:00<00:00, 17.1MB/s]"
}
},
"043e0e7fe43744589b7bad2527c2eac0": {
@@ -4324,11 +4317,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d3dab66a1c254f07afa02e73e6fd121d",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_a386fa811d524ae08ac67cce5ebf3a15",
"tabbable": null,
"tooltip": null,
- "value": "\u20071.36M/?\u2007[00:00<00:00,\u200720.0MB/s]"
+ "value": " 1.36M/? [00:00<00:00, 20.0MB/s]"
}
},
"04d1b3296c75497bb314206d6c7d5341": {
@@ -4347,11 +4340,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_8e6cf78296b14bc381f13658ebf99912",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_0ebc4d3f1e94415086e749f4cd41b783",
"tabbable": null,
"tooltip": null,
- "value": "\u20071.04M/?\u2007[00:00<00:00,\u200710.6MB/s]"
+ "value": " 1.04M/? [00:00<00:00, 10.6MB/s]"
}
},
"0582e71e725a4851a1905aceaa3c36ae": {
@@ -4670,11 +4663,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_2569c461e9144c4c82e856e4533449ba",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_2f332b079a3044fa8ab87f028a7b80b0",
"tabbable": null,
"tooltip": null,
- "value": "\u200748/48\u2007[01:10<00:00,\u2007\u20071.46s/it]"
+ "value": " 48/48 [01:10<00:00, 1.46s/it]"
}
},
"0fd7652c5e624ef7b2a36a0b0397f51d": {
@@ -4806,11 +4799,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_5cc6434927224f72aadd34dc0e0c2894",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_432ff9b9f3574d69b47e8272dd762923",
"tabbable": null,
"tooltip": null,
- "value": "merges.txt:\u2007"
+ "value": "merges.txt: "
}
},
"12f960167a1c417aacdd77ce3a997e35": {
@@ -4829,7 +4822,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_8dd28e04200641a9a2a4e5ed241db518",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_912fc8506a6f45e38f1573d27eff6457",
"tabbable": null,
"tooltip": null,
@@ -4876,11 +4869,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_ae5d013bec884f4b97d1852b1fb52432",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_87c101f3cc0f4553870ca9de688b9e83",
"tabbable": null,
"tooltip": null,
- "value": "vocab.json:\u2007"
+ "value": "vocab.json: "
}
},
"1775bd14b2104a078aa63991cc11ba85": {
@@ -4968,11 +4961,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_6dd11b3c888a461aaa85372b044ccd53",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_dc90adc8272e4e3e929844e2ceef149b",
"tabbable": null,
"tooltip": null,
- "value": "\u2007124/124\u2007[00:00<00:00,\u200754.7kB/s]"
+ "value": " 124/124 [00:00<00:00, 54.7kB/s]"
}
},
"180d2ba6e10e4e808eba69a8517d5080": {
@@ -4991,7 +4984,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_de109b45a18f42b5aa83f63ef683379f",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_4e44c9999b664ea9bad1b4e360ee76c7",
"tabbable": null,
"tooltip": null,
@@ -5687,7 +5680,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_59190b0bd8e74ee1bab6aea2f931856d",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_c0aa3c04c0a74717b8fc6700213bf579",
"tabbable": null,
"tooltip": null,
@@ -5797,11 +5790,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_700b88d4443848bab341b9d7b00cab54",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_b2e784a339524df682698e606959668e",
"tabbable": null,
"tooltip": null,
- "value": "tokenizer.json:\u2007"
+ "value": "tokenizer.json: "
}
},
"31a28b69348b40bfbd14a54380bfb766": {
@@ -5820,7 +5813,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_eddca0196bdf4e1eb5605e557bfe597b",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_c762da254b434ffea5b7c35e73009302",
"tabbable": null,
"tooltip": null,
@@ -6315,7 +6308,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_5ed82326612a4505a34bc16d6b0b5fa8",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_69d19b1cf82443ff8994ffd7b156921c",
"tabbable": null,
"tooltip": null,
@@ -6338,11 +6331,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_c9571f91e4894ac0ab6f9433d6dd7258",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_c08b955fa9494958bee9f565c568fc31",
"tabbable": null,
"tooltip": null,
- "value": "tokenizer.json:\u2007"
+ "value": "tokenizer.json: "
}
},
"3c377339930f49a5891caeb0639a8360": {
@@ -6562,11 +6555,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_68f6906692b447f1acec3cef5772fe5a",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_67cda37b9ae74d3484442b7d3bb19a26",
"tabbable": null,
"tooltip": null,
- "value": "merges.txt:\u2007"
+ "value": "merges.txt: "
}
},
"4000e5115c6d48d687ab9b9695a0d826": {
@@ -6601,7 +6594,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d94d7a4f5ab34fc9a4a4ee0a07764461",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_0fd7652c5e624ef7b2a36a0b0397f51d",
"tabbable": null,
"tooltip": null,
@@ -6822,11 +6815,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_212c5660764844db8cdc6e3a16099521",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7de33c6a558d40a69a85b3db9e203ae8",
"tabbable": null,
"tooltip": null,
- "value": "model.safetensors:\u2007100%"
+ "value": "model.safetensors: 100%"
}
},
"46c907a0ac31481f9147bf22e2ac5864": {
@@ -6898,11 +6891,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_91db9856db97451196e16d433896af48",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_4e68d17b4b7a45369d6917887ddd7a28",
"tabbable": null,
"tooltip": null,
- "value": "vocab.json:\u2007"
+ "value": "vocab.json: "
}
},
"49be2b480d5847a3af7835c317236280": {
@@ -7254,11 +7247,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d7e105c660824d349c4ee17006f04437",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_2848d61b3098431baa1d82fb85f469fe",
"tabbable": null,
"tooltip": null,
- "value": "\u2007548M/548M\u2007[00:18<00:00,\u200760.2MB/s]"
+ "value": " 548M/548M [00:18<00:00, 60.2MB/s]"
}
},
"54af6102260d458db54e634c9814aa6f": {
@@ -7295,11 +7288,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_6d3de9443ae74b75930d8397e9c7ed9a",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_63c8f7fdcf6346d9b3ca140d0f63f8e4",
"tabbable": null,
"tooltip": null,
- "value": "\u2007144/144\u2007[00:07<00:00,\u200718.78it/s]"
+ "value": " 144/144 [00:07<00:00, 18.78it/s]"
}
},
"54daaf323029464b9b67f8a4f53b3002": {
@@ -7334,7 +7327,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_39f8ccf31f6c49c7a95c59989236a3cf",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_54af6102260d458db54e634c9814aa6f",
"tabbable": null,
"tooltip": null,
@@ -7654,11 +7647,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_7ffa006c49564aa8ad58f08f48b98955",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_357dcb56edba4564b6ec3051f3e977a5",
"tabbable": null,
"tooltip": null,
- "value": "\u2007144/144\u2007[00:08<00:00,\u200715.78it/s]"
+ "value": " 144/144 [00:08<00:00, 15.78it/s]"
}
},
"5f730b9ef10e4b8bb97c4fef1bd7cbb2": {
@@ -7790,7 +7783,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_4b378e2fe92a4bb5a0cc2adee8a9372d",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_8e4b9fcabfbc4a37a54f08a99e28b220",
"tabbable": null,
"tooltip": null,
@@ -7839,11 +7832,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d7699d95a0ab4240bfa2754ac81a4dea",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_b92108f127ec4341af59d110b5f991c4",
"tabbable": null,
"tooltip": null,
- "value": "Loading\u2007weights:\u2007100%"
+ "value": "Loading weights: 100%"
}
},
"61ffe9ae7ab44e94b5729cae80e49437": {
@@ -7999,11 +7992,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_38e4ab89d9ed4f16a2a481054a18977f",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_3673315e0fe741048d2ba304360be671",
"tabbable": null,
"tooltip": null,
- "value": "Loading\u2007weights:\u2007100%"
+ "value": "Loading weights: 100%"
}
},
"67cda37b9ae74d3484442b7d3bb19a26": {
@@ -8204,11 +8197,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_e3abfde7cfd24e938684e179059edd9d",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_293230277eac481e84ab200e7cd5bdc1",
"tabbable": null,
"tooltip": null,
- "value": "\u200726.0/26.0\u2007[00:00<00:00,\u20073.64kB/s]"
+ "value": " 26.0/26.0 [00:00<00:00, 3.64kB/s]"
}
},
"6ca82062740348f2bee12629de7f8e2f": {
@@ -8351,11 +8344,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_39409ff188e6463bab5bf783828cdbd6",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7bd58f2c5b444ed9b5f21c6b364c9dce",
"tabbable": null,
"tooltip": null,
- "value": "\u2007456k/?\u2007[00:00<00:00,\u20079.34MB/s]"
+ "value": " 456k/? [00:00<00:00, 9.34MB/s]"
}
},
"6dd11b3c888a461aaa85372b044ccd53": {
@@ -8551,11 +8544,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_c08dfc48a3574ac1ba2e416960d1d3ea",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_3d29acbe122346388e66e0741971a810",
"tabbable": null,
"tooltip": null,
- "value": "generation_config.json:\u2007100%"
+ "value": "generation_config.json: 100%"
}
},
"737d22cc16184d6a92cf045c476c7a01": {
@@ -8574,11 +8567,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_8f5f5df1b0314449a113f7bb959fa273",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7671d9fee96d4e9f921046a1fb092672",
"tabbable": null,
"tooltip": null,
- "value": "\u2007689/689\u2007[00:00<00:00,\u2007131kB/s]"
+ "value": " 689/689 [00:00<00:00, 131kB/s]"
}
},
"73cd80c764784b4197af01198ba6b886": {
@@ -8597,7 +8590,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_f796802e863c48138fdcec92f546a372",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_22e77d7546334e038b64cc2c856a6a13",
"tabbable": null,
"tooltip": null,
@@ -8620,11 +8613,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_cb86f659e5ad4c5296c32b97d99d357c",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_8dacdd0b8b5e4433ae4511433eb7df1d",
"tabbable": null,
"tooltip": null,
- "value": "generation_config.json:\u2007100%"
+ "value": "generation_config.json: 100%"
}
},
"741e4c3c5c56426c91955f6b0622f629": {
@@ -8643,11 +8636,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_5e617862d88745c792f610f6651662f6",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_60e5aae936bb41faad191cf2155f78d3",
"tabbable": null,
"tooltip": null,
- "value": "tokenizer_config.json:\u2007100%"
+ "value": "tokenizer_config.json: 100%"
}
},
"75b76bb0ff2f491a8e5febeb8166cbd2": {
@@ -8814,11 +8807,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_07149da010c5489696b03653df25cdd2",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_8ccfc9585b794ba293cbe376311c42ba",
"tabbable": null,
"tooltip": null,
- "value": "\u2007144/144\u2007[00:07<00:00,\u200717.62it/s]"
+ "value": " 144/144 [00:07<00:00, 17.62it/s]"
}
},
"775073e43a7a4b2f970c810d2a05c73e": {
@@ -9415,11 +9408,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_cfee0a94009c461dbdca5b73961f7fbe",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_f732f9391dd14bcaace6fd5c27a8335a",
"tabbable": null,
"tooltip": null,
- "value": "\u2007144/144\u2007[00:07<00:00,\u200718.66it/s]"
+ "value": " 144/144 [00:07<00:00, 18.66it/s]"
}
},
"84867a129d0043b4910ac244e8a984df": {
@@ -10090,11 +10083,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_a996f052d0ed4d938f0c71fc72e8c1b6",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_2a6fd373a6524eb3ae033ea78d3cb61e",
"tabbable": null,
"tooltip": null,
- "value": "config.json:\u2007100%"
+ "value": "config.json: 100%"
}
},
"9157c241f7064a8596e1ffaeb850e59c": {
@@ -10206,7 +10199,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_c51496d64234439ebbfec98b59f44803",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_29bf4e8f0e6042b498c6ac50d8fedf68",
"tabbable": null,
"tooltip": null,
@@ -10509,11 +10502,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_d95b7b9cf6914acb9d1152502d2ba41b",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_d7c69daa5fa44a6487f5dc66380ec31a",
"tabbable": null,
"tooltip": null,
- "value": "\u2007180/180\u2007[00:10<00:00,\u200718.55it/s]"
+ "value": " 180/180 [00:10<00:00, 18.55it/s]"
}
},
"9b50599c4a084d9eb5b8755040c9bd32": {
@@ -10548,11 +10541,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_f5bbf314d840422b9e486386de3f5bb6",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_dc68cf2bf3e94df881377a106754a350",
"tabbable": null,
"tooltip": null,
- "value": "\u20071.04M/?\u2007[00:00<00:00,\u200710.5MB/s]"
+ "value": " 1.04M/? [00:00<00:00, 10.5MB/s]"
}
},
"9ba87779605c4969bf48b4071a94c630": {
@@ -10672,11 +10665,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_363c2c2e96624875af87c420c7e2cf95",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_b732ea0e03674d4384ac0d2dbf2a5f69",
"tabbable": null,
"tooltip": null,
- "value": "config.json:\u2007100%"
+ "value": "config.json: 100%"
}
},
"a0df7ae7fcc1441a8c9cca5a80b539b0": {
@@ -10695,11 +10688,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_bcaf8ebca41240bd86dde0c617b68f01",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_dd6bf89931a64c63bbcd2cf526835c2d",
"tabbable": null,
"tooltip": null,
- "value": "\u20072160/2160\u2007[02:00<00:00,\u200718.64it/s]"
+ "value": " 2160/2160 [02:00<00:00, 18.64it/s]"
}
},
"a1f007e8fb68491daa3ba444ca49f505": {
@@ -10810,11 +10803,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_7655d70259dd4dbc8bd4d288f8850b7c",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_eb49abf5e8ea41e69c1307f78fda4a90",
"tabbable": null,
"tooltip": null,
- "value": "\u2007180/180\u2007[00:09<00:00,\u200718.56it/s]"
+ "value": " 180/180 [00:09<00:00, 18.56it/s]"
}
},
"a711143026bc46a5b1b7bc3dccca1850": {
@@ -10859,11 +10852,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_2970477ecd6545b2bb698748d4019dac",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7c2906dcc3d34ae6846332eb5375cc58",
"tabbable": null,
"tooltip": null,
- "value": "\u2007124/124\u2007[00:00<00:00,\u200773.2kB/s]"
+ "value": " 124/124 [00:00<00:00, 73.2kB/s]"
}
},
"a92811787eb84dd19d9ec2fb2eab7eee": {
@@ -11382,11 +11375,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_b75438a235a243d49404467d54b63373",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_8f8ffe07f8314800ad637194c8c7d10f",
"tabbable": null,
"tooltip": null,
- "value": "\u2007665/665\u2007[00:00<00:00,\u2007122kB/s]"
+ "value": " 665/665 [00:00<00:00, 122kB/s]"
}
},
"b59d1c8089e04592a1b87a7c198d1f6c": {
@@ -11482,7 +11475,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_97f6ad1918334a3ab503d4a5da11c9ef",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_0e453235a18e4f5c9008040b1420f718",
"tabbable": null,
"tooltip": null,
@@ -11610,11 +11603,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_45e840f42f8c46d491678fad7820e835",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_e29dd892780a4208b61593998e588e1f",
"tabbable": null,
"tooltip": null,
- "value": "model.safetensors:\u2007100%"
+ "value": "model.safetensors: 100%"
}
},
"bb91ef19e455404aac3d283f868f9687": {
@@ -12051,11 +12044,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_2452dd39a79742b29964500360f4a478",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_6ca82062740348f2bee12629de7f8e2f",
"tabbable": null,
"tooltip": null,
- "value": "\u2007580/580\u2007[00:00<00:00,\u20075679.21it/s,\u2007Materializing\u2007param=transformer.wte.weight]"
+ "value": " 580/580 [00:00<00:00, 5679.21it/s, Materializing param=transformer.wte.weight]"
}
},
"c455478a557645b29777950e364a5006": {
@@ -12259,11 +12252,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_cb791fd8c7ae49079f9162125e98ff79",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_52cd22d9c796437692165d3f3ed48e82",
"tabbable": null,
"tooltip": null,
- "value": "\u20072160/2160\u2007[02:00<00:00,\u200715.47it/s]"
+ "value": " 2160/2160 [02:00<00:00, 15.47it/s]"
}
},
"c5bca44eefb940d39fba70d4fff71571": {
@@ -12359,11 +12352,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_f44f01a296fb4d198718574f1f802ba2",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_c6b1b39c22564a59bc5b950d7a46708f",
"tabbable": null,
"tooltip": null,
- "value": "\u200726.0/26.0\u2007[00:00<00:00,\u20074.24kB/s]"
+ "value": " 26.0/26.0 [00:00<00:00, 4.24kB/s]"
}
},
"c6b1b39c22564a59bc5b950d7a46708f": {
@@ -12418,11 +12411,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_7d0ba24e89554742a562ce50135447f0",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_502e8cbf7b7c47b480b8a85405fc24cf",
"tabbable": null,
"tooltip": null,
- "value": "\u20072160/2160\u2007[02:00<00:00,\u200717.95it/s]"
+ "value": " 2160/2160 [02:00<00:00, 17.95it/s]"
}
},
"c836782daaa847248009a626db347182": {
@@ -12563,11 +12556,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_3d83725b10254139a51cb688a495459d",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_020cf001eb7d496295a325cbc0ee8718",
"tabbable": null,
"tooltip": null,
- "value": "\u20072160/2160\u2007[02:00<00:00,\u200718.75it/s]"
+ "value": " 2160/2160 [02:00<00:00, 18.75it/s]"
}
},
"cac816368dee481a9fee6b196a2b16d6": {
@@ -12761,7 +12754,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_72035846c38c4a439583cc5e974c0a52",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_7aebe1ba464949849da0e182d90d0669",
"tabbable": null,
"tooltip": null,
@@ -12987,11 +12980,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_8ab4ec47b46e401883c185417c452f17",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_fdb4d541f8414fe4bee9265554cd7522",
"tabbable": null,
"tooltip": null,
- "value": "tokenizer_config.json:\u2007100%"
+ "value": "tokenizer_config.json: 100%"
}
},
"d3dab66a1c254f07afa02e73e6fd121d": {
@@ -13693,11 +13686,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_b41e54c58689400b86fc0dbf18e4bbaa",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_98bce027ddf046c29bbbb03c6e9b1de3",
"tabbable": null,
"tooltip": null,
- "value": "\u2007144/144\u2007[00:08<00:00,\u200718.70it/s]"
+ "value": " 144/144 [00:08<00:00, 18.70it/s]"
}
},
"df2f48a5055b4cb7a9db8115c407fd8c": {
@@ -13893,11 +13886,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_29376b858cd4489a8fcefc2b096df1e5",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_6f7575088f10441c87ded2f68ac37e9f",
"tabbable": null,
"tooltip": null,
- "value": "\u20072160/2160\u2007[01:59<00:00,\u200718.72it/s]"
+ "value": " 2160/2160 [01:59<00:00, 18.72it/s]"
}
},
"e3abfde7cfd24e938684e179059edd9d": {
@@ -14220,11 +14213,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_db91637067cb428c94674237eabda8f7",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_dd98180b8cba40aa82fc6221dd4676d0",
"tabbable": null,
"tooltip": null,
- "value": "\u2007180/180\u2007[00:09<00:00,\u200717.32it/s]"
+ "value": " 180/180 [00:09<00:00, 17.32it/s]"
}
},
"eb49abf5e8ea41e69c1307f78fda4a90": {
@@ -14765,7 +14758,7 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_e2c844c07e434e718186afaf72312371",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_3704e0be72444fd9a07038fbe1b19156",
"tabbable": null,
"tooltip": null,
@@ -14841,11 +14834,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_61ffe9ae7ab44e94b5729cae80e49437",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_a314648aea3d40f89ac931c47f81c9e6",
"tabbable": null,
"tooltip": null,
- "value": "\u2007148/148\u2007[00:00<00:00,\u20075498.19it/s,\u2007Materializing\u2007param=transformer.wte.weight]"
+ "value": " 148/148 [00:00<00:00, 5498.19it/s, Materializing param=transformer.wte.weight]"
}
},
"fd150a5176074e959dfa52a35770b5f0": {
@@ -14864,11 +14857,11 @@
"description": "",
"description_allow_html": false,
"layout": "IPY_MODEL_a05625d67e634f2a80c65cdcfcbe8f8c",
- "placeholder": "\u200b",
+ "placeholder": "",
"style": "IPY_MODEL_d1e752f79bbc40ddbae7a02895c9b74e",
"tabbable": null,
"tooltip": null,
- "value": "\u20076.43G/6.43G\u2007[04:52<00:00,\u2007111MB/s]"
+ "value": " 6.43G/6.43G [04:52<00:00, 111MB/s]"
}
},
"fdb4d541f8414fe4bee9265554cd7522": {
@@ -14950,4 +14943,4 @@
},
"nbformat": 4,
"nbformat_minor": 2
-}
\ No newline at end of file
+}
diff --git a/demos/LLaMA.ipynb b/demos/LLaMA.ipynb
index 24e0efd48..faf1fed52 100644
--- a/demos/LLaMA.ipynb
+++ b/demos/LLaMA.ipynb
@@ -27,7 +27,7 @@
},
{
"cell_type": "code",
- "execution_count": 1,
+ "execution_count": 11,
"metadata": {},
"outputs": [
{
@@ -35,6 +35,8 @@
"output_type": "stream",
"text": [
"Running as a Jupyter notebook - intended for development only!\n",
+ "The autoreload extension is already loaded. To reload it, use:\n",
+ " %reload_ext autoreload\n",
"Using renderer: colab\n"
]
}
@@ -81,10 +83,12 @@
},
{
"cell_type": "code",
- "execution_count": 2,
+ "execution_count": null,
"metadata": {},
"outputs": [],
"source": [
+ "# NBVAL_IGNORE_OUTPUT\n",
+ "\n",
"# Import stuff\n",
"import torch\n",
"import tqdm.auto as tqdm\n",
@@ -101,7 +105,6 @@
") # Hooking utilities\n",
"from transformer_lens.model_bridge import TransformerBridge\n",
"\n",
- "# NBVAL_IGNORE_OUTPUT\n",
"_ = torch.set_grad_enabled(False)\n",
"\n",
"def imshow(tensor, renderer=None, xaxis=\"\", yaxis=\"\", **kwargs):\n",
@@ -154,7 +157,7 @@
},
{
"cell_type": "code",
- "execution_count": 3,
+ "execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
@@ -179,8 +182,8 @@
"cell_type": "markdown",
"metadata": {},
"source": [
- "## Loading LLaMA-2\n",
- "LLaMA-2 is hosted on HuggingFace, but gated by login.\n",
+ "## Loading LLaMA-3.2\n",
+ "LLaMA-3.2 is hosted on HuggingFace, but gated by login.\n",
"\n",
"Before running the notebook, either set your `HF_TOKEN` environment variable or log in via the CLI:\n",
"```bash\n",
@@ -191,18 +194,18 @@
},
{
"cell_type": "code",
- "execution_count": 4,
+ "execution_count": 14,
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
- "model_id": "b60d7b0399ed41ebbfbc6bc4244b05ee",
+ "model_id": "6613d10ad8974b08bb07b4a1294c43d4",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
- "Loading weights: 0%| | 0/291 [00:00\n",
+ "\n",
" "
],
"text/plain": [
- ""
+ ""
]
},
"metadata": {},
@@ -355,23 +358,23 @@
},
{
"cell_type": "code",
- "execution_count": 7,
+ "execution_count": 19,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Shape of the value tensor: torch.Size([1, 34, 32, 128])\n",
- "Original Loss: 2.930\n",
- "Ablated Loss: 2.879\n"
+ "Shape of the value tensor: torch.Size([1, 32, 8, 64])\n",
+ "Original Loss: 2.615\n",
+ "Ablated Loss: 2.598\n"
]
}
],
"source": [
"# NBVAL_IGNORE_OUTPUT\n",
"layer_to_ablate = 0\n",
- "head_index_to_ablate = 31\n",
+ "head_index_to_ablate = 4\n",
"\n",
"# We define a head ablation hook\n",
"# The type annotations are NOT necessary, they're just a useful guide to the reader\n",
diff --git a/demos/LLaMA2_GPU_Quantized.ipynb b/demos/LLaMA2_GPU_Quantized.ipynb
index 56aa07925..3129e6489 100644
--- a/demos/LLaMA2_GPU_Quantized.ipynb
+++ b/demos/LLaMA2_GPU_Quantized.ipynb
@@ -6,7 +6,7 @@
"id": "EyASOtpeCUsO"
},
"source": [
- "# LLaMA and Llama-2 in TransformerLens"
+ "# LLaMA in TransformerLens (Quantized)"
]
},
{
@@ -20,7 +20,7 @@
},
{
"cell_type": "code",
- "execution_count": 6,
+ "execution_count": 18,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
@@ -74,7 +74,7 @@
},
{
"cell_type": "code",
- "execution_count": 7,
+ "execution_count": 19,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
@@ -105,12 +105,14 @@
},
{
"cell_type": "code",
- "execution_count": 9,
+ "execution_count": null,
"metadata": {
"id": "P8zS3MPkCUsR"
},
"outputs": [],
"source": [
+ "# NBVAL_IGNORE_OUTPUT\n",
+ "\n",
"# Import stuff\n",
"import torch\n",
"import tqdm.auto as tqdm\n",
@@ -127,7 +129,6 @@
") # Hooking utilities\n",
"from transformer_lens.model_bridge import TransformerBridge\n",
"\n",
- "# NBVAL_IGNORE_OUTPUT\n",
"_ = torch.set_grad_enabled(False)\n",
"\n",
"def imshow(tensor, renderer=None, xaxis=\"\", yaxis=\"\", **kwargs):\n",
@@ -185,7 +186,7 @@
},
{
"cell_type": "code",
- "execution_count": 10,
+ "execution_count": 21,
"metadata": {
"id": "RdJ0AuW_CUsS"
},
@@ -214,8 +215,8 @@
"id": "UmOqXE9wCUsS"
},
"source": [
- "## Loading LLaMA-2\n",
- "LLaMA-2 is hosted on HuggingFace, but gated by login.\n",
+ "## Loading LLaMA-3.2\n",
+ "LLaMA-3.2 is hosted on HuggingFace, but gated by login.\n",
"\n",
"Before running the notebook, either set your `HF_TOKEN` environment variable or log in via the CLI:\n",
"```bash\n",
@@ -252,7 +253,7 @@
},
{
"cell_type": "code",
- "execution_count": 17,
+ "execution_count": 22,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
@@ -388,12 +389,96 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
- "model_id": "bb6704de83c640d3ac1b5d4046b23414",
+ "model_id": "02974f818bc54305b535861303ca208e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
- "Loading weights: 0%| | 0/291 [00:00\n",
+ "\n",
" "
],
"text/plain": [
- ""
+ ""
]
},
"metadata": {},
@@ -682,7 +747,7 @@
},
{
"cell_type": "code",
- "execution_count": 16,
+ "execution_count": 31,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
@@ -695,16 +760,16 @@
"name": "stdout",
"output_type": "stream",
"text": [
- "Shape of the value tensor: torch.Size([1, 34, 32, 128])\n",
- "Original Loss: 2.957\n",
- "Ablated Loss: 2.918\n"
+ "Shape of the value tensor: torch.Size([1, 32, 8, 64])\n",
+ "Original Loss: 2.951\n",
+ "Ablated Loss: 2.934\n"
]
}
],
"source": [
"# NBVAL_IGNORE_OUTPUT\n",
"layer_to_ablate = 0\n",
- "head_index_to_ablate = 31\n",
+ "head_index_to_ablate = 4\n",
"\n",
"# We define a head ablation hook\n",
"# The type annotations are NOT necessary, they're just a useful guide to the reader\n",
diff --git a/demos/stable_lm.ipynb b/demos/stable_lm.ipynb
index 419b2cbfc..077952edb 100644
--- a/demos/stable_lm.ipynb
+++ b/demos/stable_lm.ipynb
@@ -29,7 +29,7 @@
},
{
"cell_type": "code",
- "execution_count": 21,
+ "execution_count": 1,
"metadata": {
"id": "D_OSNfeCSjS2"
},
@@ -38,9 +38,7 @@
"name": "stdout",
"output_type": "stream",
"text": [
- "Running as a Jupyter notebook - intended for development only!\n",
- "The autoreload extension is already loaded. To reload it, use:\n",
- " %reload_ext autoreload\n"
+ "Running as a Jupyter notebook - intended for development only!\n"
]
}
],
@@ -65,7 +63,7 @@
},
{
"cell_type": "code",
- "execution_count": 22,
+ "execution_count": 2,
"metadata": {
"id": "PXB6xkimoH2h"
},
@@ -80,7 +78,7 @@
},
{
"cell_type": "code",
- "execution_count": 23,
+ "execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
@@ -205,12 +203,12 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
- "model_id": "07d32635e7f444b8b8591b25fc39f141",
+ "model_id": "db184d0e866c489b8753520f96b2d771",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
- "Loading weights: 0%| | 0/196 [00:00