Text Generation
Transformers
Safetensors
mistral
svd
compression
safety
interpretability
conversational
text-generation-inference
Instructions to use Jeesup/svd-safety-mis7_swift_remove50 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Jeesup/svd-safety-mis7_swift_remove50 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Jeesup/svd-safety-mis7_swift_remove50") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Jeesup/svd-safety-mis7_swift_remove50") model = AutoModelForCausalLM.from_pretrained("Jeesup/svd-safety-mis7_swift_remove50", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Jeesup/svd-safety-mis7_swift_remove50 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Jeesup/svd-safety-mis7_swift_remove50" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-mis7_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove50
- SGLang
How to use Jeesup/svd-safety-mis7_swift_remove50 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Jeesup/svd-safety-mis7_swift_remove50" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-mis7_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Jeesup/svd-safety-mis7_swift_remove50" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Jeesup/svd-safety-mis7_swift_remove50", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Jeesup/svd-safety-mis7_swift_remove50 with Docker Model Runner:
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove50
Download compression.json from Jeesup/svd-safety-mis7_swift_remove50: direct link, hf CLI and curl.
- Browser
- Download file 15.1 kB
-
https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove50/resolve/main/compression.json
- Command line
-
hf download hf://Jeesup/svd-safety-mis7_swift_remove50/compression.json
-
curl -L -o compression.json https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove50/resolve/main/compression.json
15.1 kB
| { | |
| "source": "/workspace/svd_safety/assets/models/Mistral-7B-Instruct-v0.2", | |
| "upstream": "hiahei/Swift-SVD (compression) + AIoT-MLSys-Lab/SVD-LLM utils/LoRA.py (recovery)", | |
| "stage": "finetuned", | |
| "ratio_removed": 0.5, | |
| "internal_ratio_kept": 0.5, | |
| "from_pt": "/workspace/svd_safety/results/swift_ft/cells/mis7_swift_remove50/finetune_v/merge.pt", | |
| "mean_rel_err": 0.5888538884464651, | |
| "max_rel_err": 0.7068923115730286, | |
| "param_fraction": 0.5002723106971154, | |
| "ranks": { | |
| "0.q_proj": 1152, | |
| "0.k_proj": 485, | |
| "0.v_proj": 442, | |
| "0.o_proj": 1052, | |
| "0.gate_proj": 1680, | |
| "0.up_proj": 1680, | |
| "0.down_proj": 1702, | |
| "1.q_proj": 1157, | |
| "1.k_proj": 493, | |
| "1.v_proj": 465, | |
| "1.o_proj": 1084, | |
| "1.gate_proj": 1680, | |
| "1.up_proj": 1680, | |
| "1.down_proj": 1976, | |
| "2.q_proj": 942, | |
| "2.k_proj": 374, | |
| "2.v_proj": 367, | |
| "2.o_proj": 907, | |
| "2.gate_proj": 1680, | |
| "2.up_proj": 1680, | |
| "2.down_proj": 1351, | |
| "3.q_proj": 940, | |
| "3.k_proj": 372, | |
| "3.v_proj": 363, | |
| "3.o_proj": 898, | |
| "3.gate_proj": 1680, | |
| "3.up_proj": 1680, | |
| "3.down_proj": 1353, | |
| "4.q_proj": 938, | |
| "4.k_proj": 371, | |
| "4.v_proj": 366, | |
| "4.o_proj": 903, | |
| "4.gate_proj": 1680, | |
| "4.up_proj": 1680, | |
| "4.down_proj": 1371, | |
| "5.q_proj": 938, | |
| "5.k_proj": 372, | |
| "5.v_proj": 366, | |
| "5.o_proj": 910, | |
| "5.gate_proj": 1680, | |
| "5.up_proj": 1680, | |
| "5.down_proj": 1384, | |
| "6.q_proj": 939, | |
| "6.k_proj": 373, | |
| "6.v_proj": 368, | |
| "6.o_proj": 917, | |
| "6.gate_proj": 1680, | |
| "6.up_proj": 1680, | |
| "6.down_proj": 1395, | |
| "7.q_proj": 938, | |
| "7.k_proj": 373, | |
| "7.v_proj": 368, | |
| "7.o_proj": 925, | |
| "7.gate_proj": 1680, | |
| "7.up_proj": 1680, | |
| "7.down_proj": 1406, | |
| "8.q_proj": 935, | |
| "8.k_proj": 371, | |
| "8.v_proj": 368, | |
| "8.o_proj": 924, | |
| "8.gate_proj": 1680, | |
| "8.up_proj": 1680, | |
| "8.down_proj": 1403, | |
| "9.q_proj": 934, | |
| "9.k_proj": 370, | |
| "9.v_proj": 369, | |
| "9.o_proj": 924, | |
| "9.gate_proj": 1680, | |
| "9.up_proj": 1680, | |
| "9.down_proj": 1409, | |
| "10.q_proj": 935, | |
| "10.k_proj": 371, | |
| "10.v_proj": 369, | |
| "10.o_proj": 932, | |
| "10.gate_proj": 1680, | |
| "10.up_proj": 1680, | |
| "10.down_proj": 1410, | |
| "11.q_proj": 933, | |
| "11.k_proj": 370, | |
| "11.v_proj": 370, | |
| "11.o_proj": 934, | |
| "11.gate_proj": 1680, | |
| "11.up_proj": 1680, | |
| "11.down_proj": 1412, | |
| "12.q_proj": 939, | |
| "12.k_proj": 374, | |
| "12.v_proj": 373, | |
| "12.o_proj": 941, | |
| "12.gate_proj": 1680, | |
| "12.up_proj": 1680, | |
| "12.down_proj": 1428, | |
| "13.q_proj": 936, | |
| "13.k_proj": 372, | |
| "13.v_proj": 372, | |
| "13.o_proj": 939, | |
| "13.gate_proj": 1680, | |
| "13.up_proj": 1680, | |
| "13.down_proj": 1428, | |
| "14.q_proj": 944, | |
| "14.k_proj": 376, | |
| "14.v_proj": 378, | |
| "14.o_proj": 949, | |
| "14.gate_proj": 1680, | |
| "14.up_proj": 1680, | |
| "14.down_proj": 1451, | |
| "15.q_proj": 955, | |
| "15.k_proj": 383, | |
| "15.v_proj": 385, | |
| "15.o_proj": 964, | |
| "15.gate_proj": 1680, | |
| "15.up_proj": 1680, | |
| "15.down_proj": 1481, | |
| "16.q_proj": 959, | |
| "16.k_proj": 385, | |
| "16.v_proj": 387, | |
| "16.o_proj": 968, | |
| "16.gate_proj": 1680, | |
| "16.up_proj": 1680, | |
| "16.down_proj": 1500, | |
| "17.q_proj": 978, | |
| "17.k_proj": 395, | |
| "17.v_proj": 397, | |
| "17.o_proj": 990, | |
| "17.gate_proj": 1680, | |
| "17.up_proj": 1680, | |
| "17.down_proj": 1543, | |
| "18.q_proj": 978, | |
| "18.k_proj": 394, | |
| "18.v_proj": 398, | |
| "18.o_proj": 993, | |
| "18.gate_proj": 1680, | |
| "18.up_proj": 1680, | |
| "18.down_proj": 1555, | |
| "19.q_proj": 965, | |
| "19.k_proj": 388, | |
| "19.v_proj": 391, | |
| "19.o_proj": 983, | |
| "19.gate_proj": 1680, | |
| "19.up_proj": 1680, | |
| "19.down_proj": 1537, | |
| "20.q_proj": 942, | |
| "20.k_proj": 375, | |
| "20.v_proj": 379, | |
| "20.o_proj": 955, | |
| "20.gate_proj": 1680, | |
| "20.up_proj": 1680, | |
| "20.down_proj": 1488, | |
| "21.q_proj": 936, | |
| "21.k_proj": 371, | |
| "21.v_proj": 376, | |
| "21.o_proj": 947, | |
| "21.gate_proj": 1680, | |
| "21.up_proj": 1680, | |
| "21.down_proj": 1477, | |
| "22.q_proj": 926, | |
| "22.k_proj": 365, | |
| "22.v_proj": 371, | |
| "22.o_proj": 934, | |
| "22.gate_proj": 1680, | |
| "22.up_proj": 1680, | |
| "22.down_proj": 1457, | |
| "23.q_proj": 927, | |
| "23.k_proj": 365, | |
| "23.v_proj": 371, | |
| "23.o_proj": 940, | |
| "23.gate_proj": 1680, | |
| "23.up_proj": 1680, | |
| "23.down_proj": 1460, | |
| "24.q_proj": 926, | |
| "24.k_proj": 365, | |
| "24.v_proj": 371, | |
| "24.o_proj": 944, | |
| "24.gate_proj": 1680, | |
| "24.up_proj": 1680, | |
| "24.down_proj": 1461, | |
| "25.q_proj": 929, | |
| "25.k_proj": 366, | |
| "25.v_proj": 374, | |
| "25.o_proj": 951, | |
| "25.gate_proj": 1680, | |
| "25.up_proj": 1680, | |
| "25.down_proj": 1471, | |
| "26.q_proj": 933, | |
| "26.k_proj": 368, | |
| "26.v_proj": 377, | |
| "26.o_proj": 957, | |
| "26.gate_proj": 1680, | |
| "26.up_proj": 1680, | |
| "26.down_proj": 1483, | |
| "27.q_proj": 936, | |
| "27.k_proj": 370, | |
| "27.v_proj": 379, | |
| "27.o_proj": 958, | |
| "27.gate_proj": 1680, | |
| "27.up_proj": 1680, | |
| "27.down_proj": 1493, | |
| "28.q_proj": 949, | |
| "28.k_proj": 378, | |
| "28.v_proj": 387, | |
| "28.o_proj": 976, | |
| "28.gate_proj": 1680, | |
| "28.up_proj": 1680, | |
| "28.down_proj": 1532, | |
| "29.q_proj": 957, | |
| "29.k_proj": 381, | |
| "29.v_proj": 392, | |
| "29.o_proj": 994, | |
| "29.gate_proj": 1680, | |
| "29.up_proj": 1680, | |
| "29.down_proj": 1550, | |
| "30.q_proj": 971, | |
| "30.k_proj": 389, | |
| "30.v_proj": 401, | |
| "30.o_proj": 1013, | |
| "30.gate_proj": 1680, | |
| "30.up_proj": 1680, | |
| "30.down_proj": 1593, | |
| "31.q_proj": 1053, | |
| "31.k_proj": 433, | |
| "31.v_proj": 448, | |
| "31.o_proj": 1114, | |
| "31.gate_proj": 1680, | |
| "31.up_proj": 1680, | |
| "31.down_proj": 1816 | |
| }, | |
| "rel_err": { | |
| "0.q_proj": 0.10284276306629181, | |
| "0.k_proj": 0.09011625498533249, | |
| "0.v_proj": 0.4146844744682312, | |
| "0.o_proj": 0.4441743791103363, | |
| "0.gate_proj": 0.4323641359806061, | |
| "0.up_proj": 0.4812891185283661, | |
| "0.down_proj": 0.6020375490188599, | |
| "1.q_proj": 0.1907663345336914, | |
| "1.k_proj": 0.13933585584163666, | |
| "1.v_proj": 0.5992577075958252, | |
| "1.o_proj": 0.5850644111633301, | |
| "1.gate_proj": 0.5018978714942932, | |
| "1.up_proj": 0.5401790142059326, | |
| "1.down_proj": 0.5769923329353333, | |
| "2.q_proj": 0.5321782827377319, | |
| "2.k_proj": 0.5425702929496765, | |
| "2.v_proj": 0.682027280330658, | |
| "2.o_proj": 0.6367513537406921, | |
| "2.gate_proj": 0.5818910598754883, | |
| "2.up_proj": 0.6077799201011658, | |
| "2.down_proj": 0.681037187576294, | |
| "3.q_proj": 0.48435133695602417, | |
| "3.k_proj": 0.4475176930427551, | |
| "3.v_proj": 0.7068923115730286, | |
| "3.o_proj": 0.677703857421875, | |
| "3.gate_proj": 0.5896062254905701, | |
| "3.up_proj": 0.6225286722183228, | |
| "3.down_proj": 0.6739526987075806, | |
| "4.q_proj": 0.4879390597343445, | |
| "4.k_proj": 0.4703967571258545, | |
| "4.v_proj": 0.6710249185562134, | |
| "4.o_proj": 0.6244789361953735, | |
| "4.gate_proj": 0.5910874605178833, | |
| "4.up_proj": 0.6321871876716614, | |
| "4.down_proj": 0.6802555918693542, | |
| "5.q_proj": 0.5210317969322205, | |
| "5.k_proj": 0.5119411945343018, | |
| "5.v_proj": 0.6989429593086243, | |
| "5.o_proj": 0.6632655262947083, | |
| "5.gate_proj": 0.5846965312957764, | |
| "5.up_proj": 0.6342030763626099, | |
| "5.down_proj": 0.6687344908714294, | |
| "6.q_proj": 0.505556583404541, | |
| "6.k_proj": 0.48692795634269714, | |
| "6.v_proj": 0.6872745752334595, | |
| "6.o_proj": 0.646428108215332, | |
| "6.gate_proj": 0.5995981693267822, | |
| "6.up_proj": 0.6370501518249512, | |
| "6.down_proj": 0.6577861309051514, | |
| "7.q_proj": 0.4832315444946289, | |
| "7.k_proj": 0.4549109935760498, | |
| "7.v_proj": 0.6812716722488403, | |
| "7.o_proj": 0.6252717971801758, | |
| "7.gate_proj": 0.5871505737304688, | |
| "7.up_proj": 0.6294843554496765, | |
| "7.down_proj": 0.654327392578125, | |
| "8.q_proj": 0.49038824439048767, | |
| "8.k_proj": 0.4526974856853485, | |
| "8.v_proj": 0.6452953219413757, | |
| "8.o_proj": 0.6178920269012451, | |
| "8.gate_proj": 0.5910043716430664, | |
| "8.up_proj": 0.6263056993484497, | |
| "8.down_proj": 0.647315502166748, | |
| "9.q_proj": 0.48501771688461304, | |
| "9.k_proj": 0.4705408811569214, | |
| "9.v_proj": 0.6292407512664795, | |
| "9.o_proj": 0.6174561977386475, | |
| "9.gate_proj": 0.5865891575813293, | |
| "9.up_proj": 0.6233162879943848, | |
| "9.down_proj": 0.6318148970603943, | |
| "10.q_proj": 0.4836317300796509, | |
| "10.k_proj": 0.4697504937648773, | |
| "10.v_proj": 0.6263429522514343, | |
| "10.o_proj": 0.6223848462104797, | |
| "10.gate_proj": 0.5854211449623108, | |
| "10.up_proj": 0.6197803020477295, | |
| "10.down_proj": 0.6372787356376648, | |
| "11.q_proj": 0.4799323081970215, | |
| "11.k_proj": 0.4386899471282959, | |
| "11.v_proj": 0.6085149049758911, | |
| "11.o_proj": 0.6194866895675659, | |
| "11.gate_proj": 0.585040271282196, | |
| "11.up_proj": 0.6186556816101074, | |
| "11.down_proj": 0.6358811259269714, | |
| "12.q_proj": 0.46628451347351074, | |
| "12.k_proj": 0.4467337131500244, | |
| "12.v_proj": 0.6398075222969055, | |
| "12.o_proj": 0.602120578289032, | |
| "12.gate_proj": 0.5809478759765625, | |
| "12.up_proj": 0.6136550307273865, | |
| "12.down_proj": 0.6333811283111572, | |
| "13.q_proj": 0.49495774507522583, | |
| "13.k_proj": 0.4739055931568146, | |
| "13.v_proj": 0.6350283026695251, | |
| "13.o_proj": 0.6104384660720825, | |
| "13.gate_proj": 0.5884830951690674, | |
| "13.up_proj": 0.6139264702796936, | |
| "13.down_proj": 0.6283634305000305, | |
| "14.q_proj": 0.4639151692390442, | |
| "14.k_proj": 0.44637247920036316, | |
| "14.v_proj": 0.6475556492805481, | |
| "14.o_proj": 0.6104862093925476, | |
| "14.gate_proj": 0.5859494209289551, | |
| "14.up_proj": 0.6189398169517517, | |
| "14.down_proj": 0.6564309597015381, | |
| "15.q_proj": 0.46387919783592224, | |
| "15.k_proj": 0.42992931604385376, | |
| "15.v_proj": 0.675900399684906, | |
| "15.o_proj": 0.6175917387008667, | |
| "15.gate_proj": 0.5837653279304504, | |
| "15.up_proj": 0.6173660159111023, | |
| "15.down_proj": 0.6548573970794678, | |
| "16.q_proj": 0.4708837568759918, | |
| "16.k_proj": 0.4432294964790344, | |
| "16.v_proj": 0.6696433424949646, | |
| "16.o_proj": 0.6041324138641357, | |
| "16.gate_proj": 0.5891684293746948, | |
| "16.up_proj": 0.626923680305481, | |
| "16.down_proj": 0.656101644039154, | |
| "17.q_proj": 0.5025116801261902, | |
| "17.k_proj": 0.48880496621131897, | |
| "17.v_proj": 0.6514131426811218, | |
| "17.o_proj": 0.6207662224769592, | |
| "17.gate_proj": 0.6006857752799988, | |
| "17.up_proj": 0.6366637349128723, | |
| "17.down_proj": 0.6493645906448364, | |
| "18.q_proj": 0.4747740626335144, | |
| "18.k_proj": 0.4571532607078552, | |
| "18.v_proj": 0.6963456869125366, | |
| "18.o_proj": 0.5963596701622009, | |
| "18.gate_proj": 0.6004551649093628, | |
| "18.up_proj": 0.6350386142730713, | |
| "18.down_proj": 0.6672616004943848, | |
| "19.q_proj": 0.5268400311470032, | |
| "19.k_proj": 0.47958895564079285, | |
| "19.v_proj": 0.7042666077613831, | |
| "19.o_proj": 0.6199771165847778, | |
| "19.gate_proj": 0.6060331463813782, | |
| "19.up_proj": 0.6367842555046082, | |
| "19.down_proj": 0.6811782121658325, | |
| "20.q_proj": 0.5465222001075745, | |
| "20.k_proj": 0.5119665265083313, | |
| "20.v_proj": 0.6899744272232056, | |
| "20.o_proj": 0.6347531080245972, | |
| "20.gate_proj": 0.6125325560569763, | |
| "20.up_proj": 0.636976957321167, | |
| "20.down_proj": 0.6877227425575256, | |
| "21.q_proj": 0.5809774398803711, | |
| "21.k_proj": 0.5469288229942322, | |
| "21.v_proj": 0.6610574722290039, | |
| "21.o_proj": 0.6689409017562866, | |
| "21.gate_proj": 0.6139337420463562, | |
| "21.up_proj": 0.6391165256500244, | |
| "21.down_proj": 0.6832785606384277, | |
| "22.q_proj": 0.6217024922370911, | |
| "22.k_proj": 0.5693233609199524, | |
| "22.v_proj": 0.6632723808288574, | |
| "22.o_proj": 0.6856058835983276, | |
| "22.gate_proj": 0.6192406415939331, | |
| "22.up_proj": 0.6405629515647888, | |
| "22.down_proj": 0.6809287667274475, | |
| "23.q_proj": 0.6093561053276062, | |
| "23.k_proj": 0.5737671256065369, | |
| "23.v_proj": 0.6871448159217834, | |
| "23.o_proj": 0.6642407178878784, | |
| "23.gate_proj": 0.6219133138656616, | |
| "23.up_proj": 0.6424050331115723, | |
| "23.down_proj": 0.6742445826530457, | |
| "24.q_proj": 0.6159307360649109, | |
| "24.k_proj": 0.5868129134178162, | |
| "24.v_proj": 0.6966824531555176, | |
| "24.o_proj": 0.6866542100906372, | |
| "24.gate_proj": 0.6238306760787964, | |
| "24.up_proj": 0.6446942687034607, | |
| "24.down_proj": 0.6646919250488281, | |
| "25.q_proj": 0.6210862994194031, | |
| "25.k_proj": 0.5705580115318298, | |
| "25.v_proj": 0.679048478603363, | |
| "25.o_proj": 0.6676820516586304, | |
| "25.gate_proj": 0.6258559823036194, | |
| "25.up_proj": 0.6465917825698853, | |
| "25.down_proj": 0.6531879901885986, | |
| "26.q_proj": 0.6165540218353271, | |
| "26.k_proj": 0.5607565641403198, | |
| "26.v_proj": 0.6618245840072632, | |
| "26.o_proj": 0.6507518887519836, | |
| "26.gate_proj": 0.6270332336425781, | |
| "26.up_proj": 0.6463326811790466, | |
| "26.down_proj": 0.6390087008476257, | |
| "27.q_proj": 0.6284988522529602, | |
| "27.k_proj": 0.6069222092628479, | |
| "27.v_proj": 0.6805939674377441, | |
| "27.o_proj": 0.6412052512168884, | |
| "27.gate_proj": 0.6305440664291382, | |
| "27.up_proj": 0.6479936242103577, | |
| "27.down_proj": 0.6380788683891296, | |
| "28.q_proj": 0.5876466035842896, | |
| "28.k_proj": 0.5294674634933472, | |
| "28.v_proj": 0.6338029503822327, | |
| "28.o_proj": 0.6242429614067078, | |
| "28.gate_proj": 0.6301671266555786, | |
| "28.up_proj": 0.6454077959060669, | |
| "28.down_proj": 0.6476228833198547, | |
| "29.q_proj": 0.5711690783500671, | |
| "29.k_proj": 0.5322142243385315, | |
| "29.v_proj": 0.670903205871582, | |
| "29.o_proj": 0.6061508655548096, | |
| "29.gate_proj": 0.6246752142906189, | |
| "29.up_proj": 0.6391395926475525, | |
| "29.down_proj": 0.6555037498474121, | |
| "30.q_proj": 0.5478631854057312, | |
| "30.k_proj": 0.4874037206172943, | |
| "30.v_proj": 0.6046014428138733, | |
| "30.o_proj": 0.5960355997085571, | |
| "30.gate_proj": 0.6242106556892395, | |
| "30.up_proj": 0.6300148367881775, | |
| "30.down_proj": 0.6422404646873474, | |
| "31.q_proj": 0.49975186586380005, | |
| "31.k_proj": 0.44580137729644775, | |
| "31.v_proj": 0.5610269904136658, | |
| "31.o_proj": 0.5210819840431213, | |
| "31.gate_proj": 0.6289830803871155, | |
| "31.up_proj": 0.6037687659263611, | |
| "31.down_proj": 0.6013435125350952 | |
| }, | |
| "step": "swift_svd_dynamic_finetuned", | |
| "compression": { | |
| "upstream": "hiahei/Swift-SVD", | |
| "step": "swift_svd_dynamic", | |
| "alloc": "dynamic", | |
| "alpha": 0.6, | |
| "nsamples": 256, | |
| "seqlen": 2048, | |
| "dataset": "WikiText2", | |
| "seed": 42, | |
| "gqa_budget": "shape", | |
| "param_fraction": 0.5002723106971154, | |
| "mean_rel_err": 0.5875181926946554, | |
| "max_rel_err": 0.7061887383460999 | |
| }, | |
| "factorisation": { | |
| "max_rel_err": 0.0014601770089939237, | |
| "factor_params": 3491561472, | |
| "dense_params": 6979321856 | |
| }, | |
| "finetune": { | |
| "protocol": "svdllm_stage2_u_then_v", | |
| "script": "gitclones/SVD-LLM/utils/LoRA.py", | |
| "data": "alpaca-cleaned", | |
| "epochs_per_half": 2, | |
| "learning_rate": 0.0001, | |
| "batch_size": 64, | |
| "micro_batch_size": 4, | |
| "lora_r": 8, | |
| "lora_alpha": 16, | |
| "lora_dropout": 0.05, | |
| "cutoff_len": 256, | |
| "val_set_size": 2000, | |
| "max_train_samples": "all", | |
| "precision": "bf16 (upstream hard-codes fp16; see run_with_torch_compat)" | |
| } | |
| } |