p6e7p7/rag
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
f8f94b947e
rag/rag.ipynb

998 lines
67 KiB
Plaintext
Raw Normal View History

20240103
2024-01-03 02:46:11 -05:00
{
"cells": [
{
"cell_type": "code",
"execution_count": 43,
"id": "6503d5b0-6cf3-42d8-982c-353eb42d9d26",
"metadata": {},
"outputs": [],
"source": [
"from llama_index import (\n",
" SimpleDirectoryReader,\n",
" VectorStoreIndex,\n",
" ServiceContext,\n",
" set_global_tokenizer,\n",
" load_index_from_storage,\n",
")\n",
"from llama_index.llms import LlamaCPP\n",
"from llama_index.llms.llama_utils import (\n",
" messages_to_prompt,\n",
" completion_to_prompt,\n",
")\n",
"from llama_index.vector_stores import ChromaVectorStore\n",
"from llama_index.storage.storage_context import StorageContext\n",
"from llama_index.embeddings import HuggingFaceEmbedding\n",
"from llama_index.query_engine import CitationQueryEngine\n",
"from llama_index.prompts import PromptTemplate\n",
"\n",
"from IPython.display import Markdown, display\n",
"\n",
"from transformers import AutoTokenizer\n",
"\n",
"import os\n",
"\n",
"import chromadb"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "15e2462f-5d5b-4f4b-8119-de782c400d78",
"metadata": {},
"outputs": [],
"source": [
"def messages_to_prompt(messages):\n",
" prompt = \"\"\n",
" for message in messages:\n",
" if message.role == 'system':\n",
" prompt += f\"<|system|>\\n{message.content}</s>\\n\"\n",
" elif message.role == 'user':\n",
" prompt += f\"<|user|>\\n{message.content}</s>\\n\"\n",
" elif message.role == 'assistant':\n",
" prompt += f\"<|assistant|>\\n{message.content}</s>\\n\"\n",
"\n",
" # ensure we start with a system prompt, insert blank if needed\n",
" if not prompt.startswith(\"<|system|>\\n\"):\n",
" prompt = \"<|system|>\\n</s>\\n\" + prompt\n",
"\n",
" # add final assistant prompt\n",
" prompt = prompt + \"<|assistant|>\\n\"\n",
"\n",
" return prompt"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "373e13fb-9233-4b0b-8dc3-a1e24117bd76",
"metadata": {},
"outputs": [],
"source": [
"def completion_to_prompt(completion, system_prompt=None):\n",
" prompt = \"\"\n",
" system_prompt_str = system_prompt or \"\"\"\\\n",
"You carefully provide accurate, factual, thoughtful, nuanced answers, and are brilliant at reasoning. \\\n",
"If you think there might not be a correct answer, you say so. \\\n",
"Don't be verbose in your answers, but do provide details and examples where it might help the explanation. \\\n",
"\"\"\"\n",
" prompt = f\"<|system|>\\n {system_prompt_str.strip()} </s>\\n\"\n",
" prompt += f\"<|user|>\\n {completion} </s>\\n\"\n",
" prompt += f\"<|assistant|>\\n\"\n",
" return prompt"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "05f54aec-f155-4f4b-9b19-874c13db87cc",
"metadata": {
"collapsed": true,
"jupyter": {
"outputs_hidden": true
}
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"llama_model_loader: loaded meta data with 21 key-value pairs and 291 tensors from /Users/peportier/llm/a/a/zephyr-7b-beta.Q5_K_M.gguf (version GGUF V3 (latest))\n",
"llama_model_loader: - tensor 0: token_embd.weight q5_K [ 4096, 32000, 1, 1 ]\n",
"llama_model_loader: - tensor 1: blk.0.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 2: blk.0.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 3: blk.0.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 4: blk.0.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 5: blk.0.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 6: blk.0.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 7: blk.0.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 8: blk.0.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 9: blk.0.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 10: blk.1.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 11: blk.1.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 12: blk.1.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 13: blk.1.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 14: blk.1.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 15: blk.1.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 16: blk.1.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 17: blk.1.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 18: blk.1.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 19: blk.2.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 20: blk.2.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 21: blk.2.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 22: blk.2.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 23: blk.2.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 24: blk.2.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 25: blk.2.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 26: blk.2.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 27: blk.2.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 28: blk.3.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 29: blk.3.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 30: blk.3.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 31: blk.3.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 32: blk.3.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 33: blk.3.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 34: blk.3.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 35: blk.3.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 36: blk.3.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 37: blk.4.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 38: blk.4.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 39: blk.4.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 40: blk.4.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 41: blk.4.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 42: blk.4.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 43: blk.4.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 44: blk.4.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 45: blk.4.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 46: blk.5.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 47: blk.5.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 48: blk.5.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 49: blk.5.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 50: blk.5.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 51: blk.5.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 52: blk.5.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 53: blk.5.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 54: blk.5.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 55: blk.6.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 56: blk.6.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 57: blk.6.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 58: blk.6.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 59: blk.6.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 60: blk.6.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 61: blk.6.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 62: blk.6.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 63: blk.6.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 64: blk.7.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 65: blk.7.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 66: blk.7.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 67: blk.7.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 68: blk.7.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 69: blk.7.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 70: blk.7.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 71: blk.7.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 72: blk.7.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 73: blk.8.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 74: blk.8.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 75: blk.8.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 76: blk.8.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 77: blk.10.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 78: blk.10.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 79: blk.10.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 80: blk.10.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 81: blk.10.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 82: blk.10.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 83: blk.10.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 84: blk.10.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 85: blk.10.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 86: blk.11.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 87: blk.11.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 88: blk.11.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 89: blk.11.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 90: blk.11.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 91: blk.11.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 92: blk.11.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 93: blk.11.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 94: blk.11.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 95: blk.12.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 96: blk.12.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 97: blk.12.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 98: blk.12.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 99: blk.12.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 100: blk.12.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 101: blk.8.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 102: blk.8.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 103: blk.8.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 104: blk.8.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 105: blk.8.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 106: blk.9.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 107: blk.9.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 108: blk.9.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 109: blk.9.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 110: blk.9.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 111: blk.9.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 112: blk.9.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 113: blk.9.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 114: blk.9.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 115: blk.12.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 116: blk.12.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 117: blk.12.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 118: blk.13.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 119: blk.13.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 120: blk.13.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 121: blk.13.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 122: blk.13.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 123: blk.13.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 124: blk.13.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 125: blk.13.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 126: blk.13.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 127: blk.14.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 128: blk.14.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 129: blk.14.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 130: blk.14.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 131: blk.14.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 132: blk.14.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 133: blk.14.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 134: blk.14.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 135: blk.14.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 136: blk.15.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 137: blk.15.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 138: blk.15.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 139: blk.15.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 140: blk.15.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 141: blk.15.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 142: blk.15.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 143: blk.15.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 144: blk.15.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 145: blk.16.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 146: blk.16.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 147: blk.16.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 148: blk.16.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 149: blk.16.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 150: blk.16.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 151: blk.16.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 152: blk.16.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 153: blk.16.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 154: blk.17.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 155: blk.17.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 156: blk.17.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 157: blk.17.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 158: blk.17.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 159: blk.17.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 160: blk.17.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 161: blk.17.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 162: blk.17.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 163: blk.18.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 164: blk.18.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 165: blk.18.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 166: blk.18.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 167: blk.18.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 168: blk.18.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 169: blk.18.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 170: blk.18.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 171: blk.18.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 172: blk.19.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 173: blk.19.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 174: blk.19.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 175: blk.19.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 176: blk.19.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 177: blk.19.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 178: blk.19.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 179: blk.19.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 180: blk.19.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 181: blk.20.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 182: blk.20.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 183: blk.20.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 184: blk.20.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 185: blk.20.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 186: blk.20.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 187: blk.20.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 188: blk.20.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 189: blk.20.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 190: blk.21.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 191: blk.21.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 192: blk.21.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 193: blk.21.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 194: blk.21.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 195: blk.21.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 196: blk.21.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 197: blk.21.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 198: blk.21.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 199: blk.22.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 200: blk.22.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 201: blk.22.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 202: blk.22.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 203: blk.22.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 204: blk.22.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 205: blk.22.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 206: blk.22.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 207: blk.22.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 208: blk.23.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 209: blk.23.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 210: blk.23.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 211: blk.23.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 212: blk.23.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 213: blk.23.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 214: blk.23.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 215: blk.23.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 216: blk.23.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 217: blk.24.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 218: blk.24.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 219: blk.24.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 220: blk.24.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 221: blk.24.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 222: blk.24.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 223: blk.24.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 224: blk.24.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 225: blk.24.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 226: blk.25.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 227: blk.25.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 228: blk.25.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 229: blk.25.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 230: blk.25.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 231: blk.25.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 232: blk.25.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 233: blk.25.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 234: blk.25.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 235: blk.26.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 236: blk.26.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 237: blk.26.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 238: blk.26.attn_v.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 239: blk.26.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 240: blk.26.ffn_down.weight q5_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 241: blk.26.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 242: blk.26.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 243: blk.26.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 244: blk.27.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 245: blk.27.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 246: blk.27.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 247: blk.27.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 248: blk.27.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 249: blk.27.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 250: blk.27.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 251: blk.27.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 252: blk.27.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 253: blk.28.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 254: blk.28.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 255: blk.28.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 256: blk.28.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 257: blk.28.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 258: blk.28.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 259: blk.28.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 260: blk.28.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 261: blk.28.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 262: blk.29.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 263: blk.29.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 264: blk.29.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 265: blk.29.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 266: blk.29.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 267: blk.29.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 268: blk.29.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 269: blk.29.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 270: blk.29.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 271: blk.30.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 272: blk.30.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 273: blk.30.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 274: blk.30.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 275: blk.30.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 276: blk.30.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 277: output.weight q6_K [ 4096, 32000, 1, 1 ]\n",
"llama_model_loader: - tensor 278: blk.30.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 279: blk.30.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 280: blk.30.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 281: blk.31.attn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 282: blk.31.ffn_down.weight q6_K [ 14336, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 283: blk.31.ffn_gate.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 284: blk.31.ffn_up.weight q5_K [ 4096, 14336, 1, 1 ]\n",
"llama_model_loader: - tensor 285: blk.31.ffn_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - tensor 286: blk.31.attn_k.weight q5_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 287: blk.31.attn_output.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 288: blk.31.attn_q.weight q5_K [ 4096, 4096, 1, 1 ]\n",
"llama_model_loader: - tensor 289: blk.31.attn_v.weight q6_K [ 4096, 1024, 1, 1 ]\n",
"llama_model_loader: - tensor 290: output_norm.weight f32 [ 4096, 1, 1, 1 ]\n",
"llama_model_loader: - kv 0: general.architecture str = llama\n",
"llama_model_loader: - kv 1: general.name str = huggingfaceh4_zephyr-7b-beta\n",
"llama_model_loader: - kv 2: llama.context_length u32 = 32768\n",
"llama_model_loader: - kv 3: llama.embedding_length u32 = 4096\n",
"llama_model_loader: - kv 4: llama.block_count u32 = 32\n",
"llama_model_loader: - kv 5: llama.feed_forward_length u32 = 14336\n",
"llama_model_loader: - kv 6: llama.rope.dimension_count u32 = 128\n",
"llama_model_loader: - kv 7: llama.attention.head_count u32 = 32\n",
"llama_model_loader: - kv 8: llama.attention.head_count_kv u32 = 8\n",
"llama_model_loader: - kv 9: llama.attention.layer_norm_rms_epsilon f32 = 0.000010\n",
"llama_model_loader: - kv 10: llama.rope.freq_base f32 = 10000.000000\n",
"llama_model_loader: - kv 11: general.file_type u32 = 17\n",
"llama_model_loader: - kv 12: tokenizer.ggml.model str = llama\n",
"llama_model_loader: - kv 13: tokenizer.ggml.tokens arr[str,32000] = [\"<unk>\", \"<s>\", \"</s>\", \"<0x00>\", \"<...\n",
"llama_model_loader: - kv 14: tokenizer.ggml.scores arr[f32,32000] = [0.000000, 0.000000, 0.000000, 0.0000...\n",
"llama_model_loader: - kv 15: tokenizer.ggml.token_type arr[i32,32000] = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...\n",
"llama_model_loader: - kv 16: tokenizer.ggml.bos_token_id u32 = 1\n",
"llama_model_loader: - kv 17: tokenizer.ggml.eos_token_id u32 = 2\n",
"llama_model_loader: - kv 18: tokenizer.ggml.unknown_token_id u32 = 0\n",
"llama_model_loader: - kv 19: tokenizer.ggml.padding_token_id u32 = 2\n",
"llama_model_loader: - kv 20: general.quantization_version u32 = 2\n",
"llama_model_loader: - type f32: 65 tensors\n",
"llama_model_loader: - type q5_K: 193 tensors\n",
"llama_model_loader: - type q6_K: 33 tensors\n",
"llm_load_vocab: special tokens definition check successful ( 259/32000 ).\n",
"llm_load_print_meta: format = GGUF V3 (latest)\n",
"llm_load_print_meta: arch = llama\n",
"llm_load_print_meta: vocab type = SPM\n",
"llm_load_print_meta: n_vocab = 32000\n",
"llm_load_print_meta: n_merges = 0\n",
"llm_load_print_meta: n_ctx_train = 32768\n",
"llm_load_print_meta: n_embd = 4096\n",
"llm_load_print_meta: n_head = 32\n",
"llm_load_print_meta: n_head_kv = 8\n",
"llm_load_print_meta: n_layer = 32\n",
"llm_load_print_meta: n_rot = 128\n",
"llm_load_print_meta: n_gqa = 4\n",
"llm_load_print_meta: f_norm_eps = 0.0e+00\n",
"llm_load_print_meta: f_norm_rms_eps = 1.0e-05\n",
"llm_load_print_meta: f_clamp_kqv = 0.0e+00\n",
"llm_load_print_meta: f_max_alibi_bias = 0.0e+00\n",
"llm_load_print_meta: n_ff = 14336\n",
"llm_load_print_meta: rope scaling = linear\n",
"llm_load_print_meta: freq_base_train = 10000.0\n",
"llm_load_print_meta: freq_scale_train = 1\n",
"llm_load_print_meta: n_yarn_orig_ctx = 32768\n",
"llm_load_print_meta: rope_finetuned = unknown\n",
"llm_load_print_meta: model type = 7B\n",
"llm_load_print_meta: model ftype = mostly Q5_K - Medium\n",
"llm_load_print_meta: model params = 7.24 B\n",
"llm_load_print_meta: model size = 4.78 GiB (5.67 BPW) \n",
"llm_load_print_meta: general.name = huggingfaceh4_zephyr-7b-beta\n",
"llm_load_print_meta: BOS token = 1 '<s>'\n",
"llm_load_print_meta: EOS token = 2 '</s>'\n",
"llm_load_print_meta: UNK token = 0 '<unk>'\n",
"llm_load_print_meta: PAD token = 2 '</s>'\n",
"llm_load_print_meta: LF token = 13 '<0x0A>'\n",
"llm_load_tensors: ggml ctx size = 0.11 MiB\n",
"llm_load_tensors: mem required = 4893.10 MiB\n",
"...................................................................................................\n",
"llama_new_context_with_model: n_ctx = 3900\n",
"llama_new_context_with_model: freq_base = 10000.0\n",
"llama_new_context_with_model: freq_scale = 1\n",
"llama_new_context_with_model: kv self size = 487.50 MiB\n",
"llama_build_graph: non-view tensors processed: 740/740\n",
"ggml_metal_init: allocating\n",
"ggml_metal_init: found device: Apple M2 Max\n",
"ggml_metal_init: picking default device: Apple M2 Max\n",
"ggml_metal_init: default.metallib not found, loading from source\n",
"ggml_metal_init: loading '/Users/peportier/miniforge3/envs/RAG_ENV/lib/python3.9/site-packages/llama_cpp/ggml-metal.metal'\n",
"ggml_metal_init: GPU name: Apple M2 Max\n",
"ggml_metal_init: GPU family: MTLGPUFamilyApple8 (1008)\n",
"ggml_metal_init: hasUnifiedMemory = true\n",
"ggml_metal_init: recommendedMaxWorkingSetSize = 49152.00 MiB\n",
"ggml_metal_init: maxTransferRate = built-in GPU\n",
"llama_new_context_with_model: compute buffer total size = 278.43 MiB\n",
"llama_new_context_with_model: max tensor size = 102.54 MiB\n",
"ggml_metal_add_buffer: allocated 'data ' buffer, size = 4893.70 MiB, ( 4894.20 / 49152.00)\n",
"ggml_metal_add_buffer: allocated 'kv ' buffer, size = 487.52 MiB, ( 5381.72 / 49152.00)\n",
"ggml_metal_add_buffer: allocated 'alloc ' buffer, size = 275.38 MiB, ( 5657.09 / 49152.00)\n"
]
}
],
"source": [
"llm = LlamaCPP(\n",
" # You can pass in the URL to a GGML model to download it automatically\n",
" model_url=None,\n",
" # optionally, you can set the path to a pre-downloaded model instead of model_url\n",
" # model_path='/Users/peportier/llm/a/a/mistral-7b-openorca.Q4_K_M.gguf',\n",
" model_path='/Users/peportier/llm/a/a/zephyr-7b-beta.Q5_K_M.gguf',\n",
" temperature=0.1,\n",
" max_new_tokens=1024,\n",
" # llama2 has a context window of 4096 tokens, but we set it lower to allow for some wiggle room\n",
" context_window=3900,\n",
" # kwargs to pass to __call__()\n",
" # https://llama-cpp-python.readthedocs.io/en/latest/api-reference/#llama_cpp.Llama.__call__\n",
" generate_kwargs={\n",
" \"temperature\": 0.1,\n",
" \"mirostat_mode\": 2,\n",
" },\n",
" # kwargs to pass to __init__()\n",
" # https://llama-cpp-python.readthedocs.io/en/latest/api-reference/#llama_cpp.Llama.__init__\n",
" # set to at least 1 to use GPU\n",
" model_kwargs={\n",
" \"n_gpu_layers\": 1,\n",
" },\n",
" # transform inputs into Llama2 format\n",
" messages_to_prompt=messages_to_prompt,\n",
" completion_to_prompt=completion_to_prompt,\n",
" verbose=False,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "42734468-f0c4-46ff-8676-98d4bde99a86",
"metadata": {},
"outputs": [],
"source": [
"model_name = \"HuggingFaceH4/zephyr-7b-beta\"\n",
"set_global_tokenizer(\n",
" AutoTokenizer.from_pretrained(model_name).encode\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "39fadac2-a30e-41ca-85b5-35fc49f9842c",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"/Users/peportier/miniforge3/envs/RAG_ENV/lib/python3.9/site-packages/transformers/utils/generic.py:309: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead.\n",
" _torch_pytree._register_pytree_node(\n"
]
}
],
"source": [
"# embed_model = HuggingFaceEmbedding(model_name=\"BAAI/bge-small-en-v1.5\")\n",
"# embed_model = HuggingFaceEmbedding(model_name=\"sentence-transformers/all-MiniLM-L6-v2\")\n",
"embed_model = HuggingFaceEmbedding(model_name=\"sentence-transformers/distiluse-base-multilingual-cased-v1\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "fbb8f6f3-9175-4f90-bd67-be8efb37ae17",
"metadata": {},
"outputs": [],
"source": [
"service_context = ServiceContext.from_defaults(\n",
" llm=llm,\n",
" embed_model=embed_model,\n",
" chunk_size=512,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "c06f2c57-8f7b-41ad-9053-bdfefc80fd98",
"metadata": {},
"outputs": [],
"source": [
"if not os.path.exists(\"./index\"):\n",
" documents = SimpleDirectoryReader(\"./docs/env\").load_data()\n",
" index = VectorStoreIndex.from_documents(\n",
" documents, service_context=service_context\n",
" )\n",
" index.storage_context.persist(persist_dir=\"./index\")\n",
"else:\n",
" index = load_index_from_storage(\n",
" StorageContext.from_defaults(persist_dir=\"./index\"),\n",
" service_context=service_context,\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b583c0c0-85a1-4384-83de-0001c9452a3f",
"metadata": {},
"outputs": [],
"source": [
"# db = chromadb.PersistentClient(path=\"./chroma_db\")\n",
"# chroma_collection = db.get_or_create_collection(\"env\")\n",
"# vector_store = ChromaVectorStore(chroma_collection=chroma_collection)\n",
"# storage_context = StorageContext.from_defaults(vector_store=vector_store)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "a7a6a9f9-a3d6-47fe-8cd8-13384e07f16a",
"metadata": {},
"outputs": [],
"source": [
"# # Create chroma index\n",
"# index = VectorStoreIndex.from_documents(\n",
"# documents, storage_context=storage_context,\n",
"# service_context=service_context\n",
"# )"
]
},
{
"cell_type": "code",
"execution_count": 54,
"id": "176a1285-bb3c-46a4-892a-1957ccb74c99",
"metadata": {},
"outputs": [],
"source": [
"text_qa_template_str_fr = (\n",
" \"<|system|>: Vous êtes un assistant IA qui répond à la question posée à la fin en utilisant le contexte suivant. Toutes les informations factuelles que vous utilisez pour répondre proviennent exclusivement du contexte. Si vous ne connaissez pas la réponse, dites simplement que vous ne savez pas, n'essayez pas d'inventer une réponse. Veuillez répondre exclusivement en français. </s>\\n\"\n",
" \"<|user|>: {context_str}\\n\"\n",
" \"Question: {query_str} </s>\\n\"\n",
" \"<|assistant|>:\"\n",
")\n",
"text_qa_template_str_en = (\n",
" \"<|system|>: You are an AI assistant who answers the question posed at the end using the following context. All the factual information you use to answer comes exclusively from the context, do not use prior knowledge. If you don't know the answer, just say you don't know, don't try to invent an answer. </s>\\n\"\n",
" \"<|user|>: {context_str}\\n\"\n",
" \"Question: {query_str} </s>\\n\"\n",
" \"<|assistant|>:\"\n",
")\n",
"text_qa_template = PromptTemplate(text_qa_template_str_en)\n",
"\n",
"refine_template_str_en = (\n",
" \"<|user|>: The original query is as follows: {query_str}\\n\"\n",
" \"We have provided an existing answer: {existing_answer}\\n\"\n",
" \"We have the opportunity to refine the existing answer (only if needed) with some more context below.\\n\"\n",
" \"------------\\n\"\n",
" \"{context_msg}\\n\"\n",
" \"------------\\n\"\n",
" \"Given the new context, refine the original answer to better answer the query. If the context isn't useful, return the original answer. </s>\\n\"\n",
" \"<|assistant|>:\"\n",
")\n",
"refine_template = PromptTemplate(refine_template_str_en)"
]
},
{
"cell_type": "code",
"execution_count": 63,
"id": "02350391-a152-4b70-8ed4-a8a3dde96728",
"metadata": {},
"outputs": [],
"source": [
"query_engine = index.as_query_engine(\n",
" text_qa_template=text_qa_template,\n",
" refine_template=refine_template,\n",
" #response_mode=\"compact\",\n",
" response_mode=\"refine\",\n",
" similarity_top_k=5,\n",
" streaming=True)"
]
},
{
"cell_type": "code",
"execution_count": 57,
"id": "f979bd65-217d-480b-a377-5b3a58593156",
"metadata": {},
"outputs": [],
"source": [
"def display_prompt_dict(prompts_dict):\n",
" for k, p in prompts_dict.items():\n",
" text_md = f\"**Prompt Key**: {k}<br>\" f\"**Text:** <br>\"\n",
" display(Markdown(text_md))\n",
" print(p.get_template())\n",
" display(Markdown(\"<br><br>\"))"
]
},
{
"cell_type": "code",
"execution_count": 58,
"id": "37ba2a5d-5eca-43aa-ac28-b071e1cbfba4",
"metadata": {},
"outputs": [
{
"data": {
"text/markdown": [
"**Prompt Key**: response_synthesizer:text_qa_template<br>**Text:** <br>"
],
"text/plain": [
"<IPython.core.display.Markdown object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"<|system|>: You are an AI assistant who answers the question posed at the end using the following context. All the factual information you use to answer comes exclusively from the context, do not use prior knowledge. If you don't know the answer, just say you don't know, don't try to invent an answer. </s>\n",
"<|user|>: {context_str}\n",
"Question: {query_str} </s>\n",
"<|assistant|>:\n"
]
},
{
"data": {
"text/markdown": [
"<br><br>"
],
"text/plain": [
"<IPython.core.display.Markdown object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/markdown": [
"**Prompt Key**: response_synthesizer:refine_template<br>**Text:** <br>"
],
"text/plain": [
"<IPython.core.display.Markdown object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"<|user|>: The original query is as follows: {query_str}\n",
"We have provided an existing answer: {existing_answer}\n",
"We have the opportunity to refine the existing answer (only if needed) with some more context below.\n",
"------------\n",
"{context_msg}\n",
"------------\n",
"Given the new context, refine the original answer to better answer the query. If the context isn't useful, return the original answer. </s>\n",
"<|assistant|>:\n"
]
},
{
"data": {
"text/markdown": [
"<br><br>"
],
"text/plain": [
"<IPython.core.display.Markdown object>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"display_prompt_dict(query_engine.get_prompts())"
]
},
{
"cell_type": "code",
"execution_count": 67,
"id": "8bcff0b7-300b-4698-86df-f36171b1ee02",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"The green premium is a concept used to compare the additional cost of choosing a zero-carbon alternative over its fossil fuel counterpart. It helps in deciding which clean alternatives should be deployed now and where innovation is needed because the clean alternatives are not yet cheap enough. The calculation of green premiums involves various assumptions, and different well-informed people may arrive at different numbers.\n",
"\n",
"In the context provided, we can see that while electric vehicles (EVs) have become more affordable in recent years due to a significant decrease in battery costs and government incentives, they still come with a modest Green Premium compared to their gas-powered counterparts. For example, the Chevrolet Bolt EV is $14,000 more expensive than the gas-powered Malibu before any tax credits, but when accounting for factors such as maintenance costs and fuel expenses, the Bolt will cost 10 cents more per mile driven.\n",
"\n",
"It's essential to understand whether a given green technology is close to being as cheap as its fossil fuel counterpart, and for those that aren't close, thinking about how innovation might bring their prices down. While the Green Premium is an imperfect tool, it's better than no tool at all, and it helps in putting time, attention, and money to their best use.\n",
"\n",
"In summary, the green premium is a useful concept for comparing costs and making informed decisions about which zero-carbon solutions to deploy now and where innovation is needed. While EVs have become more affordable, they still come with a Green Premium, but as technology continues to improve, this gap may narrow, making zero-carbon alternatives more competitive in the market."
]
}
],
"source": [
"response = query_engine.query('''What is a green premium?''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": 70,
"id": "4586e568-5dcf-481b-8512-a27695bbbe69",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Le \"green premium\" désigne la différence de prix entre les technologies écologiques et leurs contreparties traditionnelles. Il s'agit de l'augmentation supplémentaire que les consommateurs doivent payer pour choisir des options écologiques plutôt que des solutions conventionnelles. Dans le cadre de cet document, l'auteur calcule ces différences de prix pour diverses technologies vertes et évalue si elles sont abordables suffisamment pour une adoption mondiale généralisée. L'auteur souligne l'importance d'avoir des primes basses pour permettre à tous les pays de respecter les normes environnementales sans coûts prohibitifs.\n",
"\n",
"L'exemple du carburant jet d'avion illustre le calcul direct des primes vertes, mais lorsque nous appliquons ce calcul de manière plus générale, nous rencontrons un problème : nous n'avons pas actuellement une équivalent vert direct dans tous les cas. Il n'y a pas de ciment à faible émission de carbone (au moins non encore). Comment obtenir un sens du coût d'une solution écologique dans ces cas ?\n",
"\n",
"Nous pouvons le faire en menant une expérience de pensée : \"Combien coûterait-il de retirer directement le carbone de l'atmosphère ?\" Cette idée a un nom ; c'est la capture directe de l'air (DAC). Le seul centre DAC en opération actuellement est basé en Suisse et absorbe du gaz qui aurait pu être émis par une centrale à charbon au Texas il y a 10 ans. Pour déterminer combien cela coûterait, nous avons besoin de deux données : les émissions mondiales et le coût de la capture directe de l'air.\n",
"\n",
"Le DAC est une technologie coûteuse et peu développée, mais si elle peut fonctionner à grande échelle, elle permettra de capturer du dioxyde de carbone qu'importe quand ou où il a été produit. Le centre DAC en opération actuellement, basé en Suisse, capture le gaz qui aurait pu être émis par une centrale à charbon au Texas il y a 10 ans.\n",
"\n",
"Voici les Green Premiums pour remplacer les combustibles fossiles avec des alternatives sans émission de carbone :\n",
"\n",
"- Pour le gazole, les biofuels avancés coûtent 103% de plus que le gazole traditionnel, et les electrofuels coûtent 234% de plus.\n",
"- Pour l'essence, les biofuels avancés coûtent 106% de plus que l'essence traditionnelle, et les electrofuels coûtent 237% de plus.\n",
"- Pour le carburant jet d'avion, les biofuels avancés coûtent 141% de plus que le carburant jet d'avion traditionnel, et les electrofuels coûtent 296% de plus.\n",
"\n",
"Ces Green Premiums montrent qu'il y a encore beaucoup de travail à faire pour rendre les technologies vertes abordables suffisamment pour une adoption mondiale généralisée. Cependant, l'auteur souligne l'importance d'avoir des primes basses pour permettre à tous les pays de respecter les normes environnementales sans coûts prohibitifs. Les calculs de ces Green Premiums fournissent des informations utiles aux décisions d'investissement dans le secteur de l'énergie en aidant à déterminer quelles technologies vertes devraient être priorisées pour le déploiement et où il est nécessaire de poursuivre les efforts de recherche et de développement car des options écologiques alternatives sont encore trop coûteuses.CPU times: user 12.6 s, sys: 2.07 s, total: 14.7 s\n",
"Wall time: 2min 16s\n"
]
}
],
"source": [
"%%time\n",
"response = query_engine.query('''Qu'est-ce que le \"green premium\". Réponds en français.''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": 71,
"id": "7c7f1e8b-cb21-4cc6-8d51-8e29ced1ba6a",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"65\n",
"How_to_Avoid_a_Climate_Disaster_for_9712958_compressed.pdf\n",
"64\n",
"How_to_Avoid_a_Climate_Disaster_for_9712958_compressed.pdf\n",
"66\n",
"How_to_Avoid_a_Climate_Disaster_for_9712958_compressed.pdf\n",
"67\n",
"How_to_Avoid_a_Climate_Disaster_for_9712958_compressed.pdf\n",
"150\n",
"How_to_Avoid_a_Climate_Disaster_for_9712958_compressed.pdf\n"
]
}
],
"source": [
"for node in response.source_nodes:\n",
" print(node.metadata[\"page_label\"] + \"\\n\" + node.metadata[\"file_name\"])"
]
},
{
"cell_type": "code",
"execution_count": 61,
"id": "7ead638c-fa97-4664-88b6-2c3ebaff7f91",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"A carbon tax is a policy tool that places a price on greenhouse gas emissions, specifically carbon dioxide (CO2). It aims to reduce greenhouse gas emissions by making high-carbon alternatives like traditional steel production and coal-based plastics more expensive. By doing so, governments can incentivize the adoption of low-carbon alternatives such as electric arc furnace (EAF) steel and zero-carbon plastics made from renewable sources or captured CO2. This policy tool sends a market signal to individuals and businesses to reduce greenhouse gas emissions and contributes to mitigating climate change while promoting the development of clean energy technology and breakthroughs that generate and store lots of zero-carbon electricity inexpensively. A carbon tax can also help ensure that the transition to a low-carbon economy is both effective and equitable as we pursue electrification as another way to reduce emissions. By making carbon-free things cheaper and carbon-emitting things more expensive, governments can nudge producers and consumers toward more efficient decisions and encourage innovation that reduces Green Premiums. This policy tool is a way to create an incentive for inventors to create competitive carbon-free alternatives by progressively increasing the price of carbon to reflect its true cost, which reflects the external costs borne by society due to greenhouse gas emissions. It's not about punishing people for their greenhouse gases but rather creating an incentive for more efficient decisions and innovation that reduces Green Premiums."
]
}
],
"source": [
"response = query_engine.query('''What is a carbon tax?''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": 64,
"id": "ac5d3d94-218a-466b-a8f4-ee68a90ffc68",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Critical biospheric boundaries refer to limits beyond which human activities could cause irreversible and potentially catastrophic environmental changes that threaten the stability and functioning of Earth's ecosystems. These boundaries include climate change, ocean acidification, loss of biodiversity, land use changes, freshwater consumption, introduction of novel entities (such as chemicals and genetically modified organisms), atmospheric aerosol loading, stratospheric ozone depletion, and global nitrogen and phosphorus cycles. It's crucial to maintain these boundaries within safe limits because they represent the most significant, persistent, existential concerns related to environmental change.\n",
"\n",
"However, it's essential to remain skeptical and doubtful while continuously reassessing our understanding of environmental issues as some critical problems, such as acid rain and stratospheric ozone depletion, were not yet recognized 40 years ago. This list would have been different back then because we didn't fully comprehend the extent and complexity of environmental challenges at that time. As we continue to learn more about these critical biospheric boundaries, it's crucial to remain skeptical and doubtful while continuously reassessing our understanding of environmental issues. This will help us to better address these concerns in a timely and effective manner.\n",
"\n",
"In addressing these critical biospheric boundaries, we must recognize that some of these issues have been better understood than others over time. For example, the negative effects of excessive water extraction from ancient deep aquifers and serious water pollution generated by food production, cities, and industries have become more apparent in recent years. Meanwhile, issues like acid rain and stratospheric ozone depletion were not fully recognized 40 years ago due to a lack of understanding about the extent and complexity of environmental challenges at that time. As we continue to learn more about these critical biospheric boundaries, it's crucial to remain skeptical and doubtful while continuously reassessing our understanding of environmental issues. This will help us to better address these concerns in a timely and effective manner.\n",
"\n",
"The author's interdisciplinary background in energy studies has led them to prefer scanning wide horizons over drilling ever-deeper holes. Their work, which spans various aspects of energy, history, social, economic, and political factors, advocates for moving away from extreme views. The author's latest book, \"SMIL_2022_how_the_world_really_works,\" continues this quest to understand the basic realities of the biosphere, history, and the world we have created while strongly advocating for addressing environmental concerns in a timely and effective manner."
]
}
],
"source": [
"response = query_engine.query('''What are critical biospheric boundaries?''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "c4ca60d8-e4c6-4348-ba47-386733a92c72",
"metadata": {},
"outputs": [],
"source": [
"query_engine_cit = CitationQueryEngine.from_args(\n",
" index,\n",
" similarity_top_k=3,\n",
" citation_chunk_size=512,\n",
" streaming=True,\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "d26b3519-1bd3-45ff-8ea5-2c924cbc70e3",
"metadata": {},
"outputs": [],
"source": [
"response = query_engine_cit.query('''What is a green premium?''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "ace3c7b0-46e1-4a37-abfb-02ac795fdf50",
"metadata": {},
"outputs": [],
"source": [
"len(response.source_nodes)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "d77a0f58-5cbe-49d9-8358-5294d647cf5f",
"metadata": {},
"outputs": [],
"source": [
"print(response.source_nodes[0].node.get_text())"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "4fc14351-32d5-4cdc-9c5f-3c6d38c5dfef",
"metadata": {},
"outputs": [],
"source": [
"print(response.source_nodes[0].node.metadata)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "ad77a1b3-a1e0-49ac-ba6b-0d7eb7852ffd",
"metadata": {},
"outputs": [],
"source": [
"response.response_txt"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "d1469839-9453-4e6e-b471-1103d9a18a47",
"metadata": {},
"outputs": [],
"source": [
"response = query_engine_cit.query('''How are fossil fuels related to food production?''')\n",
"response.print_response_stream()"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "fc62ed24-51c7-4c96-bf57-025f8db29270",
"metadata": {},
"outputs": [],
"source": [
"print(response.source_nodes[1].node.metadata)\n",
"print(response.source_nodes[1].node.get_text())"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "876154c4-2dbb-4170-a244-98f17e763955",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "RAG_ENV",
"language": "python",
"name": "rag_env"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.18"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Reference in New Issue Copy Permalink