wassname/discovering_latent_knowledge
1
0
Fork 0
mirror of https://github.com/wassname/discovering_latent_knowledge.git synced 2026-09-10 12:00:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

wip

Browse Source
This commit is contained in:
deep1
2023-08-11 19:25:45 +08:00
parent 7a3a9dae73
commit 73af185c6e
3 changed files with 233 additions and 2430 deletions
Download Patch File Download Diff File Expand all files Collapse all files
notebooks/023_train_prob.ipynb
+61 -2257
View File
File diff suppressed because one or more lines are too long
notebooks/03_make_dataset.ipynb
+171 -172
View File
@@ -11,7 +11,7 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
@@ -32,9 +32,20 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 2,
"metadata": {},
"outputs": [],
"outputs": [
{
"data": {
"text/plain": [
"'4.30.1'"
]
},
"execution_count": 2,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import numpy as np\n",
"\n",
@@ -62,7 +73,7 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
@@ -84,7 +95,7 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
@@ -97,7 +108,7 @@
"dataset_params = dict(\n",
" model_repo=\"HuggingFaceH4/starchat-beta\",\n",
" dataset_name = \"amazon_polarity\",\n",
" N = 509, # 8000 # 4000 in 4 hours\n",
" N = 9009, # 8000 # 4000 in 4 hours\n",
" N_SHOTS = 2,\n",
" prompt_fmt=format_guard_prompt,\n",
" choices=default_class2choices,\n",
@@ -131,7 +142,7 @@
},
{
"cell_type": "code",
"execution_count": 30,
"execution_count": 5,
"metadata": {},
"outputs": [
{
@@ -140,139 +151,102 @@
"text": [
"\u001b[1mchanging pad_token_id from None to 0\u001b[0m\n",
"\u001b[1mchanging padding_side from right to left\u001b[0m\n",
"\u001b[1mchanging truncation_side from right to left\u001b[0m\n"
"\u001b[1mchanging truncation_side from right to left\u001b[0m\n",
"/home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/bitsandbytes/cuda_setup/main.py:149: UserWarning: Found duplicate ['libcudart.so', 'libcudart.so.11.0', 'libcudart.so.12.0'] files: {PosixPath('/home/ubuntu/mambaforge/envs/dlk2/lib/libcudart.so'), PosixPath('/home/ubuntu/mambaforge/envs/dlk2/lib/libcudart.so.11.0')}.. We'll flip a coin and try one of these, in order to fail forward.\n",
"Either way, this might cause trouble in the future:\n",
"If you get `CUDA error: invalid device function` errors, the above might be the cause and the solution is to make sure only one ['libcudart.so', 'libcudart.so.11.0', 'libcudart.so.12.0'] in the paths that we search based on your env.\n",
" warn(msg)\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"===================================BUG REPORT===================================\n",
"Welcome to bitsandbytes. For bug reports, please run\n",
"\n",
"python -m bitsandbytes\n",
"\n",
" and submit this information together with your error trace to: https://github.com/TimDettmers/bitsandbytes/issues\n",
"================================================================================\n",
"bin /home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/bitsandbytes/libbitsandbytes_cuda117.so\n",
"CUDA SETUP: CUDA runtime path found: /home/ubuntu/mambaforge/envs/dlk2/lib/libcudart.so\n",
"CUDA SETUP: Highest compute capability among GPUs detected: 8.6\n",
"CUDA SETUP: Detected CUDA version 117\n",
"CUDA SETUP: Loading binary /home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/bitsandbytes/libbitsandbytes_cuda117.so...\n"
]
},
{
"data": {
"text/html": [
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #800000; text-decoration-color: #800000\">╭─────────────────────────────── </span><span style=\"color: #800000; text-decoration-color: #800000; font-weight: bold\">Traceback </span><span style=\"color: #bf7f7f; text-decoration-color: #bf7f7f; font-weight: bold\">(most recent call last)</span><span style=\"color: #800000; text-decoration-color: #800000\"> ────────────────────────────────╮</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">&lt;module&gt;</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">1</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>1 model, tokenizer = load_model(model_repo=dataset_params[<span style=\"color: #808000; text-decoration-color: #808000\">'model_repo'</span>]) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2 </span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/home/ubuntu/Documents/mjc/elk/discovering_latent_knowledge/src/models/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">load.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">23</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">load_model</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">20 </span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">21 </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">def</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">load_model</span>(model_repo = <span style=\"color: #808000; text-decoration-color: #808000\">\"HuggingFaceH4/starchat-beta\"</span>, lora_repo=<span style=\"color: #0000ff; text-decoration-color: #0000ff\">None</span>, verbose=<span style=\"color: #0000ff; text-decoration-color: #0000ff\">True</span>) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">22 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> <span style=\"color: #808000; text-decoration-color: #808000\">\"starchat\"</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> model_repo: <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>23 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ </span>model, tokenizer = load_starchat(model_repo=model_repo) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">24 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"># elif \"llama\" in model_repo:</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">25 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\"># model, tokenizer = load_llama(model_repo=model_repo, lora_repo=lora_repo)</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">26 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">else</span>: <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/home/ubuntu/Documents/mjc/elk/discovering_latent_knowledge/src/models/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">load.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">51</span> in <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #00ff00; text-decoration-color: #00ff00\">load_starchat</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">48 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span>verbose_change_param(tokenizer, <span style=\"color: #808000; text-decoration-color: #808000\">'padding_side'</span>, <span style=\"color: #808000; text-decoration-color: #808000\">'left'</span>) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">49 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span>verbose_change_param(tokenizer, <span style=\"color: #808000; text-decoration-color: #808000\">'truncation_side'</span>, <span style=\"color: #808000; text-decoration-color: #808000\">'left'</span>) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">50 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>51 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span>model = AutoModelForCausalLM.from_pretrained(model_repo, config=config, **model_opti <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">52 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">53 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">return</span> model, tokenizer <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">54 </span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/transformers/models/auto/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">auto_fact</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">ory.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">484</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">from_pretrained</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">481 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ </span>) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">482 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">elif</span> <span style=\"color: #00ffff; text-decoration-color: #00ffff\">type</span>(config) <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> <span style=\"color: #00ffff; text-decoration-color: #00ffff\">cls</span>._model_mapping.keys(): <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">483 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ </span>model_class = _get_model_class(config, <span style=\"color: #00ffff; text-decoration-color: #00ffff\">cls</span>._model_mapping) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>484 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">return</span> model_class.from_pretrained( <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">485 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ │ </span>pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">486 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ </span>) <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">487 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">raise</span> <span style=\"color: #00ffff; text-decoration-color: #00ffff\">ValueError</span>( <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">/home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/transformers/</span><span style=\"color: #808000; text-decoration-color: #808000; font-weight: bold\">modeling_utils.py</span>:<span style=\"color: #0000ff; text-decoration-color: #0000ff\">281</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #0000ff; text-decoration-color: #0000ff\">9</span> in <span style=\"color: #00ff00; text-decoration-color: #00ff00\">from_pretrained</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2816 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ │ │ </span>key: device_map[key] <span style=\"color: #0000ff; text-decoration-color: #0000ff\">for</span> key <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> device_map.keys() <span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> key <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">not</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> modu <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2817 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ │ </span>} <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2818 </span><span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ │ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">if</span> <span style=\"color: #808000; text-decoration-color: #808000\">\"cpu\"</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> device_map_without_lm_head.values() <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">or</span> <span style=\"color: #808000; text-decoration-color: #808000\">\"disk\"</span> <span style=\"color: #ff00ff; text-decoration-color: #ff00ff\">in</span> device_map_ <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #800000; text-decoration-color: #800000\">❱ </span>2819 <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">│ │ │ │ │ </span><span style=\"color: #0000ff; text-decoration-color: #0000ff\">raise</span> <span style=\"color: #00ffff; text-decoration-color: #00ffff\">ValueError</span>( <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2820 </span><span style=\"color: #bfbfbf; text-decoration-color: #bfbfbf\">│ │ │ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">\"\"\"</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2821 </span><span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">│ │ │ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">Some modules are dispatched on the CPU or the disk. Make sure yo</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">│</span> <span style=\"color: #7f7f7f; text-decoration-color: #7f7f7f\">2822 </span><span style=\"color: #bfbf7f; text-decoration-color: #bfbf7f\">│ │ │ │ │ │ </span><span style=\"color: #808000; text-decoration-color: #808000\">the quantized model. If you want to dispatch the model on the CP</span> <span style=\"color: #800000; text-decoration-color: #800000\">│</span>\n",
"<span style=\"color: #800000; text-decoration-color: #800000\">╰──────────────────────────────────────────────────────────────────────────────────────────────────╯</span>\n",
"<span style=\"color: #ff0000; text-decoration-color: #ff0000; font-weight: bold\">ValueError: </span>\n",
" Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to \n",
"fit\n",
" the quantized model. If you want to dispatch the model on the CPU or the disk while keeping\n",
" these modules in <span style=\"color: #008080; text-decoration-color: #008080; font-weight: bold\">32</span>-bit, you need to set `<span style=\"color: #808000; text-decoration-color: #808000\">load_in_8bit_fp32_cpu_offload</span>=<span style=\"color: #00ff00; text-decoration-color: #00ff00; font-style: italic\">True</span>` and pass a \n",
"custom\n",
" `device_map` to `from_pretrained`. Check\n",
" <span style=\"color: #0000ff; text-decoration-color: #0000ff; text-decoration: underline\">https://huggingface.co/docs/transformers/main/en/main_classes/quantization#offload-between-</span>\n",
"<span style=\"color: #0000ff; text-decoration-color: #0000ff; text-decoration: underline\">cpu-and-gpu</span>\n",
" for more details.\n",
" \n",
"</pre>\n"
],
"application/vnd.jupyter.widget-view+json": {
"model_id": "0c4350e56b7846fba9c6725c33de2d76",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"\u001b[31m╭─\u001b[0m\u001b[31m──────────────────────────────\u001b[0m\u001b[31m \u001b[0m\u001b[1;31mTraceback \u001b[0m\u001b[1;2;31m(most recent call last)\u001b[0m\u001b[31m \u001b[0m\u001b[31m───────────────────────────────\u001b[0m\u001b[31m─╮\u001b[0m\n",
"\u001b[31m│\u001b[0m in \u001b[92m<module>\u001b[0m:\u001b[94m1\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m1 model, tokenizer = load_model(model_repo=dataset_params[\u001b[33m'\u001b[0m\u001b[33mmodel_repo\u001b[0m\u001b[33m'\u001b[0m]) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2 \u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2;33m/home/ubuntu/Documents/mjc/elk/discovering_latent_knowledge/src/models/\u001b[0m\u001b[1;33mload.py\u001b[0m:\u001b[94m23\u001b[0m in \u001b[92mload_model\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m20 \u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m21 \u001b[0m\u001b[94mdef\u001b[0m \u001b[92mload_model\u001b[0m(model_repo = \u001b[33m\"\u001b[0m\u001b[33mHuggingFaceH4/starchat-beta\u001b[0m\u001b[33m\"\u001b[0m, lora_repo=\u001b[94mNone\u001b[0m, verbose=\u001b[94mTrue\u001b[0m) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m22 \u001b[0m\u001b[2m│ \u001b[0m\u001b[94mif\u001b[0m \u001b[33m\"\u001b[0m\u001b[33mstarchat\u001b[0m\u001b[33m\"\u001b[0m \u001b[95min\u001b[0m model_repo: \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m23 \u001b[2m│ │ \u001b[0mmodel, tokenizer = load_starchat(model_repo=model_repo) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m24 \u001b[0m\u001b[2m│ \u001b[0m\u001b[2m# elif \"llama\" in model_repo:\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m25 \u001b[0m\u001b[2m│ \u001b[0m\u001b[2m# model, tokenizer = load_llama(model_repo=model_repo, lora_repo=lora_repo)\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m26 \u001b[0m\u001b[2m│ \u001b[0m\u001b[94melse\u001b[0m: \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2;33m/home/ubuntu/Documents/mjc/elk/discovering_latent_knowledge/src/models/\u001b[0m\u001b[1;33mload.py\u001b[0m:\u001b[94m51\u001b[0m in \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[92mload_starchat\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m48 \u001b[0m\u001b[2m│ \u001b[0mverbose_change_param(tokenizer, \u001b[33m'\u001b[0m\u001b[33mpadding_side\u001b[0m\u001b[33m'\u001b[0m, \u001b[33m'\u001b[0m\u001b[33mleft\u001b[0m\u001b[33m'\u001b[0m) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m49 \u001b[0m\u001b[2m│ \u001b[0mverbose_change_param(tokenizer, \u001b[33m'\u001b[0m\u001b[33mtruncation_side\u001b[0m\u001b[33m'\u001b[0m, \u001b[33m'\u001b[0m\u001b[33mleft\u001b[0m\u001b[33m'\u001b[0m) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m50 \u001b[0m\u001b[2m│ \u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m51 \u001b[2m│ \u001b[0mmodel = AutoModelForCausalLM.from_pretrained(model_repo, config=config, **model_opti \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m52 \u001b[0m\u001b[2m│ \u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m53 \u001b[0m\u001b[2m│ \u001b[0m\u001b[94mreturn\u001b[0m model, tokenizer \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m54 \u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2;33m/home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/transformers/models/auto/\u001b[0m\u001b[1;33mauto_fact\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[1;33mory.py\u001b[0m:\u001b[94m484\u001b[0m in \u001b[92mfrom_pretrained\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m481 \u001b[0m\u001b[2m│ │ │ \u001b[0m) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m482 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94melif\u001b[0m \u001b[96mtype\u001b[0m(config) \u001b[95min\u001b[0m \u001b[96mcls\u001b[0m._model_mapping.keys(): \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m483 \u001b[0m\u001b[2m│ │ │ \u001b[0mmodel_class = _get_model_class(config, \u001b[96mcls\u001b[0m._model_mapping) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m484 \u001b[2m│ │ │ \u001b[0m\u001b[94mreturn\u001b[0m model_class.from_pretrained( \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m485 \u001b[0m\u001b[2m│ │ │ │ \u001b[0mpretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m486 \u001b[0m\u001b[2m│ │ │ \u001b[0m) \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m487 \u001b[0m\u001b[2m│ │ \u001b[0m\u001b[94mraise\u001b[0m \u001b[96mValueError\u001b[0m( \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2;33m/home/ubuntu/mambaforge/envs/dlk2/lib/python3.9/site-packages/transformers/\u001b[0m\u001b[1;33mmodeling_utils.py\u001b[0m:\u001b[94m281\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[94m9\u001b[0m in \u001b[92mfrom_pretrained\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2816 \u001b[0m\u001b[2m│ │ │ │ │ \u001b[0mkey: device_map[key] \u001b[94mfor\u001b[0m key \u001b[95min\u001b[0m device_map.keys() \u001b[94mif\u001b[0m key \u001b[95mnot\u001b[0m \u001b[95min\u001b[0m modu \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2817 \u001b[0m\u001b[2m│ │ │ │ \u001b[0m} \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2818 \u001b[0m\u001b[2m│ │ │ │ \u001b[0m\u001b[94mif\u001b[0m \u001b[33m\"\u001b[0m\u001b[33mcpu\u001b[0m\u001b[33m\"\u001b[0m \u001b[95min\u001b[0m device_map_without_lm_head.values() \u001b[95mor\u001b[0m \u001b[33m\"\u001b[0m\u001b[33mdisk\u001b[0m\u001b[33m\"\u001b[0m \u001b[95min\u001b[0m device_map_ \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[31m❱ \u001b[0m2819 \u001b[2m│ │ │ │ │ \u001b[0m\u001b[94mraise\u001b[0m \u001b[96mValueError\u001b[0m( \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2820 \u001b[0m\u001b[2;90m│ │ │ │ │ │ \u001b[0m\u001b[33m\"\"\"\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2821 \u001b[0m\u001b[2;33m│ │ │ │ │ │ \u001b[0m\u001b[33mSome modules are dispatched on the CPU or the disk. Make sure yo\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m│\u001b[0m \u001b[2m2822 \u001b[0m\u001b[2;33m│ │ │ │ │ │ \u001b[0m\u001b[33mthe quantized model. If you want to dispatch the model on the CP\u001b[0m \u001b[31m│\u001b[0m\n",
"\u001b[31m╰──────────────────────────────────────────────────────────────────────────────────────────────────╯\u001b[0m\n",
"\u001b[1;91mValueError: \u001b[0m\n",
" Some modules are dispatched on the CPU or the disk. Make sure you have enough GPU RAM to \n",
"fit\n",
" the quantized model. If you want to dispatch the model on the CPU or the disk while keeping\n",
" these modules in \u001b[1;36m32\u001b[0m-bit, you need to set `\u001b[33mload_in_8bit_fp32_cpu_offload\u001b[0m=\u001b[3;92mTrue\u001b[0m` and pass a \n",
"custom\n",
" `device_map` to `from_pretrained`. Check\n",
" \u001b[4;94mhttps://huggingface.co/docs/transformers/main/en/main_classes/quantization#offload-between-\u001b[0m\n",
"\u001b[4;94mcpu-and-gpu\u001b[0m\n",
" for more details.\n",
" \n"
"Loading checkpoint shards: 0%| | 0/4 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"GPTBigCodeConfig {\n",
" \"_name_or_path\": \"HuggingFaceH4/starchat-beta\",\n",
" \"activation_function\": \"gelu\",\n",
" \"architectures\": [\n",
" \"GPTBigCodeForCausalLM\"\n",
" ],\n",
" \"attention_softmax_in_fp32\": true,\n",
" \"attn_pdrop\": 0.1,\n",
" \"bos_token_id\": 0,\n",
" \"embd_pdrop\": 0.1,\n",
" \"eos_token_id\": 0,\n",
" \"inference_runner\": 0,\n",
" \"initializer_range\": 0.02,\n",
" \"layer_norm_epsilon\": 1e-05,\n",
" \"max_batch_size\": null,\n",
" \"max_sequence_length\": null,\n",
" \"model_type\": \"gpt_bigcode\",\n",
" \"multi_query\": true,\n",
" \"n_embd\": 6144,\n",
" \"n_head\": 48,\n",
" \"n_inner\": 24576,\n",
" \"n_layer\": 40,\n",
" \"n_positions\": 8192,\n",
" \"pad_key_length\": true,\n",
" \"pre_allocate_kv_cache\": false,\n",
" \"quantization_config\": {\n",
" \"bnb_4bit_compute_dtype\": \"float32\",\n",
" \"bnb_4bit_quant_type\": \"fp4\",\n",
" \"bnb_4bit_use_double_quant\": false,\n",
" \"llm_int8_enable_fp32_cpu_offload\": false,\n",
" \"llm_int8_has_fp16_weight\": false,\n",
" \"llm_int8_skip_modules\": null,\n",
" \"llm_int8_threshold\": 6.0,\n",
" \"load_in_4bit\": true,\n",
" \"load_in_8bit\": false\n",
" },\n",
" \"resid_pdrop\": 0.1,\n",
" \"scale_attention_softmax_in_fp32\": true,\n",
" \"scale_attn_weights\": true,\n",
" \"summary_activation\": null,\n",
" \"summary_first_dropout\": 0.1,\n",
" \"summary_proj_to_labels\": true,\n",
" \"summary_type\": \"cls_index\",\n",
" \"summary_use_proj\": true,\n",
" \"torch_dtype\": \"bfloat16\",\n",
" \"transformers_version\": \"4.30.1\",\n",
" \"use_cache\": false,\n",
" \"validate_runner_input\": true,\n",
" \"vocab_size\": 49156\n",
"}\n",
"\n"
]
}
],
"source": [
@@ -289,7 +263,7 @@
},
{
"cell_type": "code",
"execution_count": 31,
"execution_count": 6,
"metadata": {},
"outputs": [
{
@@ -302,7 +276,7 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "6ee109e341e5461b8d62bc59f2dc64b3",
"model_id": "81ee1347db5d4e18b62056c580cf1441",
"version_major": 2,
"version_minor": 0
},
@@ -328,7 +302,7 @@
"})"
]
},
"execution_count": 31,
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
@@ -353,7 +327,7 @@
},
{
"cell_type": "code",
"execution_count": 32,
"execution_count": 7,
"metadata": {},
"outputs": [
{
@@ -384,7 +358,7 @@
},
{
"cell_type": "code",
"execution_count": 33,
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
@@ -414,31 +388,64 @@
},
{
"cell_type": "code",
"execution_count": 34,
"execution_count": 9,
"metadata": {
"notebookRunGroups": {
"groupValue": ""
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Loading cached processed dataset at /home/ubuntu/.cache/huggingface/datasets/amazon_polarity/amazon_polarity/3.0.0/a27b32b7e7b88eb274a8fa8ba0f654f1fe998a87c22547557317793b5d2772dc/cache-88dc9f1fd8b901a8.arrow\n",
"Loading cached processed dataset at /home/ubuntu/.cache/huggingface/datasets/amazon_polarity/amazon_polarity/3.0.0/a27b32b7e7b88eb274a8fa8ba0f654f1fe998a87c22547557317793b5d2772dc/cache-e752326d3340220d.arrow\n",
"Loading cached processed dataset at /home/ubuntu/.cache/huggingface/datasets/amazon_polarity/amazon_polarity/3.0.0/a27b32b7e7b88eb274a8fa8ba0f654f1fe998a87c22547557317793b5d2772dc/cache-7cb61ae701a89a12.arrow\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "4550272988a24d68bc8b9376a0c81fe7",
"model_id": "42428c7a72b54c5385bc45a9c3af6a63",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/509 [00:00<?, ? examples/s]"
"Map: 0%| | 0/9009 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "38298c2818a34f1b83efeaf4830421d3",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/9009 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "af83348bd74b4afabe5ede373c533fff",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/9009 [00:00<?, ? examples/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "fdb75381552440e587c9f9ae15166938",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Map: 0%| | 0/9009 [00:00<?, ? examples/s]"
]
},
"metadata": {},
@@ -449,11 +456,11 @@
"text/plain": [
"Dataset({\n",
" features: ['label', 'title', 'content', 'text', 'prompt', 'lie', 'input_ids', 'attention_mask', 'prompt_truncated'],\n",
" num_rows: 509\n",
" num_rows: 9009\n",
"})"
]
},
"execution_count": 34,
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
@@ -491,14 +498,14 @@
},
{
"cell_type": "code",
"execution_count": 35,
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"../.ds/model-starchat-beta_ds-amazon-polarity_format-guard-prompt_N509_2shots_5c2070\n"
"../.ds/model-starchat-beta_ds-amazon-polarity_format-guard-prompt_N9009_2shots_d6813d\n"
]
}
],
@@ -513,7 +520,7 @@
},
{
"cell_type": "code",
"execution_count": 36,
"execution_count": 11,
"metadata": {},
"outputs": [
{
@@ -549,13 +556,13 @@
" 'tokenizer': GPT2TokenizerFast(name_or_path='HuggingFaceH4/starchat-beta', vocab_size=49152, model_max_length=1000000000000000019884624838656, is_fast=True, padding_side='left', truncation_side='left', special_tokens={'bos_token': '<|endoftext|>', 'eos_token': '<|endoftext|>', 'unk_token': '<|endoftext|>', 'pad_token': '<|endoftext|>', 'additional_special_tokens': ['<|system|>', '<|user|>', '<|assistant|>', '<|end|>']}, clean_up_tokenization_spaces=True),\n",
" 'data': Dataset({\n",
" features: ['label', 'title', 'content', 'text', 'prompt', 'lie', 'input_ids', 'attention_mask', 'prompt_truncated'],\n",
" num_rows: 509\n",
" num_rows: 9009\n",
" }),\n",
" 'n': 509,\n",
" 'n': 9009,\n",
" 'batch_size': 10}"
]
},
"execution_count": 36,
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
@@ -573,20 +580,20 @@
},
{
"cell_type": "code",
"execution_count": 37,
"execution_count": 12,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Downloading and preparing dataset None/../.ds/model-starchat-beta_ds-amazon-polarity_format-guard-prompt_N509_2shots_5c2070 to /home/ubuntu/.cache/huggingface/datasets/generator/default-f46bbb923bbf3943/0.0.0...\n"
"Downloading and preparing dataset None/../.ds/model-starchat-beta_ds-amazon-polarity_format-guard-prompt_N9009_2shots_d6813d to /home/ubuntu/.cache/huggingface/datasets/generator/default-4cd9d3e07734db42/0.0.0...\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "6aa11b88940248c199146d73079893d5",
"model_id": "8f625b953205443fbf9388fd506528f3",
"version_major": 2,
"version_minor": 0
},
@@ -600,12 +607,12 @@
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "f4dd1c7bebb84bb6998442b6e821df80",
"model_id": "e9d293506403423cb671cf4e98ad38e1",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"get hidden states: 0%| | 0/51 [00:00<?, ?it/s]"
"get hidden states: 0%| | 0/901 [00:00<?, ?it/s]"
]
},
"metadata": {},
@@ -761,7 +768,7 @@
"source": [
"# QC by viewing a row\n",
"r = ds4[0]\n",
"print(r['prompt_truncated'][0])\n",
"print(r['prompt_truncated'])\n",
"print(r['txt_ans1'])"
]
},
@@ -834,25 +841,17 @@
"metadata": {},
"outputs": [],
"source": [
"true_switch_sign = ds4['label'][:, 0]*2-1\n",
"true_switch_sign = ds4['true'][:, 0]*2-1\n",
"true_switch_sign = ds4['label']*2-1\n",
"# true_switch_sign = ds4['true'][:, 0]*2-1\n",
"y = ((ds4['ans1'] - ds4['ans0']) * true_switch_sign) > 0"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# n = len(df)\n",
"from sklearn.preprocessing import RobustScaler\n",
"from sklearn.linear_model import LogisticRegression\n",
"from sklearn.metrics import f1_score, roc_auc_score, accuracy_score\n",
src/datasets/load.py
+1 -1
View File
@@ -20,7 +20,7 @@ def ds2df(ds, cols=None):
if cols is None:
r = ds[0]
# get all the columns that not large lists or arrays
cols = [k for k,v in r.items() if (isinstance(v, np.ndarray) and len(v)<3) or not isinstance(v, (list, np.ndarray))]
cols = [k for k,v in r.items() if (isinstance(v, np.ndarray) and v.size<3) or not isinstance(v, (list, np.ndarray))]
df = ds.select_columns(cols)
df = pd.DataFrame([rows_item(r) for r in df])