Text Generation
Transformers
Safetensors
mistral
svd
compression
safety
interpretability
conversational
text-generation-inference
Instructions to use Jeesup/svd-safety-mis7_swift_remove30 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Jeesup/svd-safety-mis7_swift_remove30 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Jeesup/svd-safety-mis7_swift_remove30") 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_remove30") model = AutoModelForCausalLM.from_pretrained("Jeesup/svd-safety-mis7_swift_remove30", 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_remove30 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_remove30" # 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_remove30", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove30
- SGLang
How to use Jeesup/svd-safety-mis7_swift_remove30 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_remove30" \ --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_remove30", "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_remove30" \ --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_remove30", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Jeesup/svd-safety-mis7_swift_remove30 with Docker Model Runner:
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove30
Download compression.json from Jeesup/svd-safety-mis7_swift_remove30: direct link, hf CLI and curl.
- Browser
- Download file 15.2 kB
-
https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove30/resolve/main/compression.json
- Command line
-
hf download hf://Jeesup/svd-safety-mis7_swift_remove30/compression.json
-
curl -L -o compression.json https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove30/resolve/main/compression.json
15.2 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": 0.3, | |
| "internal_ratio_kept": 0.7, | |
| "from_pt": "/workspace/svd_safety/results/swift_ft/cells/mis7_swift_remove30/finetune_v/merge.pt", | |
| "mean_rel_err": 0.46171893551945686, | |
| "max_rel_err": 0.5801312327384949, | |
| "param_fraction": 0.7003361628605769, | |
| "ranks": { | |
| "0.q_proj": 1613, | |
| "0.k_proj": 679, | |
| "0.v_proj": 617, | |
| "0.o_proj": 1473, | |
| "0.gate_proj": 2352, | |
| "0.up_proj": 2352, | |
| "0.down_proj": 2382, | |
| "1.q_proj": 1619, | |
| "1.k_proj": 690, | |
| "1.v_proj": 650, | |
| "1.o_proj": 1517, | |
| "1.gate_proj": 2352, | |
| "1.up_proj": 2352, | |
| "1.down_proj": 2765, | |
| "2.q_proj": 1319, | |
| "2.k_proj": 523, | |
| "2.v_proj": 513, | |
| "2.o_proj": 1270, | |
| "2.gate_proj": 2352, | |
| "2.up_proj": 2352, | |
| "2.down_proj": 1892, | |
| "3.q_proj": 1316, | |
| "3.k_proj": 520, | |
| "3.v_proj": 508, | |
| "3.o_proj": 1257, | |
| "3.gate_proj": 2352, | |
| "3.up_proj": 2352, | |
| "3.down_proj": 1895, | |
| "4.q_proj": 1314, | |
| "4.k_proj": 520, | |
| "4.v_proj": 512, | |
| "4.o_proj": 1264, | |
| "4.gate_proj": 2352, | |
| "4.up_proj": 2352, | |
| "4.down_proj": 1919, | |
| "5.q_proj": 1313, | |
| "5.k_proj": 520, | |
| "5.v_proj": 513, | |
| "5.o_proj": 1274, | |
| "5.gate_proj": 2352, | |
| "5.up_proj": 2352, | |
| "5.down_proj": 1938, | |
| "6.q_proj": 1315, | |
| "6.k_proj": 522, | |
| "6.v_proj": 514, | |
| "6.o_proj": 1285, | |
| "6.gate_proj": 2352, | |
| "6.up_proj": 2352, | |
| "6.down_proj": 1953, | |
| "7.q_proj": 1313, | |
| "7.k_proj": 522, | |
| "7.v_proj": 515, | |
| "7.o_proj": 1295, | |
| "7.gate_proj": 2352, | |
| "7.up_proj": 2352, | |
| "7.down_proj": 1969, | |
| "8.q_proj": 1310, | |
| "8.k_proj": 519, | |
| "8.v_proj": 515, | |
| "8.o_proj": 1294, | |
| "8.gate_proj": 2352, | |
| "8.up_proj": 2352, | |
| "8.down_proj": 1965, | |
| "9.q_proj": 1308, | |
| "9.k_proj": 518, | |
| "9.v_proj": 516, | |
| "9.o_proj": 1294, | |
| "9.gate_proj": 2352, | |
| "9.up_proj": 2352, | |
| "9.down_proj": 1973, | |
| "10.q_proj": 1309, | |
| "10.k_proj": 518, | |
| "10.v_proj": 516, | |
| "10.o_proj": 1305, | |
| "10.gate_proj": 2352, | |
| "10.up_proj": 2352, | |
| "10.down_proj": 1974, | |
| "11.q_proj": 1307, | |
| "11.k_proj": 518, | |
| "11.v_proj": 518, | |
| "11.o_proj": 1308, | |
| "11.gate_proj": 2352, | |
| "11.up_proj": 2352, | |
| "11.down_proj": 1977, | |
| "12.q_proj": 1315, | |
| "12.k_proj": 523, | |
| "12.v_proj": 522, | |
| "12.o_proj": 1318, | |
| "12.gate_proj": 2352, | |
| "12.up_proj": 2352, | |
| "12.down_proj": 1999, | |
| "13.q_proj": 1311, | |
| "13.k_proj": 521, | |
| "13.v_proj": 520, | |
| "13.o_proj": 1315, | |
| "13.gate_proj": 2352, | |
| "13.up_proj": 2352, | |
| "13.down_proj": 1999, | |
| "14.q_proj": 1321, | |
| "14.k_proj": 526, | |
| "14.v_proj": 529, | |
| "14.o_proj": 1328, | |
| "14.gate_proj": 2352, | |
| "14.up_proj": 2352, | |
| "14.down_proj": 2032, | |
| "15.q_proj": 1336, | |
| "15.k_proj": 535, | |
| "15.v_proj": 538, | |
| "15.o_proj": 1350, | |
| "15.gate_proj": 2352, | |
| "15.up_proj": 2352, | |
| "15.down_proj": 2073, | |
| "16.q_proj": 1343, | |
| "16.k_proj": 538, | |
| "16.v_proj": 540, | |
| "16.o_proj": 1355, | |
| "16.gate_proj": 2352, | |
| "16.up_proj": 2352, | |
| "16.down_proj": 2099, | |
| "17.q_proj": 1370, | |
| "17.k_proj": 552, | |
| "17.v_proj": 555, | |
| "17.o_proj": 1386, | |
| "17.gate_proj": 2352, | |
| "17.up_proj": 2352, | |
| "17.down_proj": 2160, | |
| "18.q_proj": 1369, | |
| "18.k_proj": 551, | |
| "18.v_proj": 556, | |
| "18.o_proj": 1390, | |
| "18.gate_proj": 2352, | |
| "18.up_proj": 2352, | |
| "18.down_proj": 2176, | |
| "19.q_proj": 1351, | |
| "19.k_proj": 542, | |
| "19.v_proj": 547, | |
| "19.o_proj": 1376, | |
| "19.gate_proj": 2352, | |
| "19.up_proj": 2352, | |
| "19.down_proj": 2151, | |
| "20.q_proj": 1318, | |
| "20.k_proj": 524, | |
| "20.v_proj": 530, | |
| "20.o_proj": 1337, | |
| "20.gate_proj": 2352, | |
| "20.up_proj": 2352, | |
| "20.down_proj": 2083, | |
| "21.q_proj": 1310, | |
| "21.k_proj": 518, | |
| "21.v_proj": 526, | |
| "21.o_proj": 1325, | |
| "21.gate_proj": 2352, | |
| "21.up_proj": 2352, | |
| "21.down_proj": 2067, | |
| "22.q_proj": 1297, | |
| "22.k_proj": 510, | |
| "22.v_proj": 519, | |
| "22.o_proj": 1308, | |
| "22.gate_proj": 2352, | |
| "22.up_proj": 2352, | |
| "22.down_proj": 2040, | |
| "23.q_proj": 1298, | |
| "23.k_proj": 511, | |
| "23.v_proj": 519, | |
| "23.o_proj": 1317, | |
| "23.gate_proj": 2352, | |
| "23.up_proj": 2352, | |
| "23.down_proj": 2044, | |
| "24.q_proj": 1297, | |
| "24.k_proj": 510, | |
| "24.v_proj": 519, | |
| "24.o_proj": 1321, | |
| "24.gate_proj": 2352, | |
| "24.up_proj": 2352, | |
| "24.down_proj": 2045, | |
| "25.q_proj": 1300, | |
| "25.k_proj": 512, | |
| "25.v_proj": 523, | |
| "25.o_proj": 1330, | |
| "25.gate_proj": 2352, | |
| "25.up_proj": 2352, | |
| "25.down_proj": 2059, | |
| "26.q_proj": 1306, | |
| "26.k_proj": 515, | |
| "26.v_proj": 528, | |
| "26.o_proj": 1340, | |
| "26.gate_proj": 2352, | |
| "26.up_proj": 2352, | |
| "26.down_proj": 2076, | |
| "27.q_proj": 1311, | |
| "27.k_proj": 518, | |
| "27.v_proj": 530, | |
| "27.o_proj": 1341, | |
| "27.gate_proj": 2352, | |
| "27.up_proj": 2352, | |
| "27.down_proj": 2090, | |
| "28.q_proj": 1328, | |
| "28.k_proj": 528, | |
| "28.v_proj": 541, | |
| "28.o_proj": 1366, | |
| "28.gate_proj": 2352, | |
| "28.up_proj": 2352, | |
| "28.down_proj": 2144, | |
| "29.q_proj": 1339, | |
| "29.k_proj": 533, | |
| "29.v_proj": 548, | |
| "29.o_proj": 1392, | |
| "29.gate_proj": 2352, | |
| "29.up_proj": 2352, | |
| "29.down_proj": 2170, | |
| "30.q_proj": 1359, | |
| "30.k_proj": 543, | |
| "30.v_proj": 561, | |
| "30.o_proj": 1418, | |
| "30.gate_proj": 2352, | |
| "30.up_proj": 2352, | |
| "30.down_proj": 2230, | |
| "31.q_proj": 1473, | |
| "31.k_proj": 605, | |
| "31.v_proj": 626, | |
| "31.o_proj": 1559, | |
| "31.gate_proj": 2352, | |
| "31.up_proj": 2352, | |
| "31.down_proj": 2541 | |
| }, | |
| "rel_err": { | |
| "0.q_proj": 0.08783026784658432, | |
| "0.k_proj": 0.076280377805233, | |
| "0.v_proj": 0.28279638290405273, | |
| "0.o_proj": 0.3450925052165985, | |
| "0.gate_proj": 0.3139413893222809, | |
| "0.up_proj": 0.34659841656684875, | |
| "0.down_proj": 0.4610687792301178, | |
| "1.q_proj": 0.14944231510162354, | |
| "1.k_proj": 0.08638541400432587, | |
| "1.v_proj": 0.42672011256217957, | |
| "1.o_proj": 0.4614779055118561, | |
| "1.gate_proj": 0.37372663617134094, | |
| "1.up_proj": 0.4030474126338959, | |
| "1.down_proj": 0.4160686731338501, | |
| "2.q_proj": 0.40826842188835144, | |
| "2.k_proj": 0.38183024525642395, | |
| "2.v_proj": 0.5488172769546509, | |
| "2.o_proj": 0.5261715054512024, | |
| "2.gate_proj": 0.44494783878326416, | |
| "2.up_proj": 0.46598997712135315, | |
| "2.down_proj": 0.5656055808067322, | |
| "3.q_proj": 0.37075719237327576, | |
| "3.k_proj": 0.3103460371494293, | |
| "3.v_proj": 0.5801312327384949, | |
| "3.o_proj": 0.570527195930481, | |
| "3.gate_proj": 0.45506104826927185, | |
| "3.up_proj": 0.48117780685424805, | |
| "3.down_proj": 0.5562737584114075, | |
| "4.q_proj": 0.3778054118156433, | |
| "4.k_proj": 0.33330559730529785, | |
| "4.v_proj": 0.5392099618911743, | |
| "4.o_proj": 0.5162532329559326, | |
| "4.gate_proj": 0.4575117528438568, | |
| "4.up_proj": 0.49015605449676514, | |
| "4.down_proj": 0.5629571676254272, | |
| "5.q_proj": 0.4073948264122009, | |
| "5.k_proj": 0.3724959194660187, | |
| "5.v_proj": 0.5715569853782654, | |
| "5.o_proj": 0.549954354763031, | |
| "5.gate_proj": 0.45217984914779663, | |
| "5.up_proj": 0.49232661724090576, | |
| "5.down_proj": 0.5492331981658936, | |
| "6.q_proj": 0.39479202032089233, | |
| "6.k_proj": 0.3444615304470062, | |
| "6.v_proj": 0.5562271475791931, | |
| "6.o_proj": 0.5352969169616699, | |
| "6.gate_proj": 0.46358194947242737, | |
| "6.up_proj": 0.49562251567840576, | |
| "6.down_proj": 0.5394559502601624, | |
| "7.q_proj": 0.3770744204521179, | |
| "7.k_proj": 0.32297107577323914, | |
| "7.v_proj": 0.5488503575325012, | |
| "7.o_proj": 0.5085676908493042, | |
| "7.gate_proj": 0.4547872841358185, | |
| "7.up_proj": 0.48921388387680054, | |
| "7.down_proj": 0.5321547985076904, | |
| "8.q_proj": 0.3834538757801056, | |
| "8.k_proj": 0.32134145498275757, | |
| "8.v_proj": 0.5088310837745667, | |
| "8.o_proj": 0.4981762766838074, | |
| "8.gate_proj": 0.4569913148880005, | |
| "8.up_proj": 0.48666849732398987, | |
| "8.down_proj": 0.5265653729438782, | |
| "9.q_proj": 0.3789723515510559, | |
| "9.k_proj": 0.3307971656322479, | |
| "9.v_proj": 0.4929361045360565, | |
| "9.o_proj": 0.5009838342666626, | |
| "9.gate_proj": 0.453980952501297, | |
| "9.up_proj": 0.4856830835342407, | |
| "9.down_proj": 0.5116001963615417, | |
| "10.q_proj": 0.3785707354545593, | |
| "10.k_proj": 0.3373726010322571, | |
| "10.v_proj": 0.4934564232826233, | |
| "10.o_proj": 0.5008266568183899, | |
| "10.gate_proj": 0.4539565443992615, | |
| "10.up_proj": 0.48363447189331055, | |
| "10.down_proj": 0.5184407234191895, | |
| "11.q_proj": 0.3743845522403717, | |
| "11.k_proj": 0.3172908425331116, | |
| "11.v_proj": 0.465146541595459, | |
| "11.o_proj": 0.5011467337608337, | |
| "11.gate_proj": 0.4532534182071686, | |
| "11.up_proj": 0.48237940669059753, | |
| "11.down_proj": 0.516564130783081, | |
| "12.q_proj": 0.36303064227104187, | |
| "12.k_proj": 0.30660074949264526, | |
| "12.v_proj": 0.5052470564842224, | |
| "12.o_proj": 0.4829132556915283, | |
| "12.gate_proj": 0.44975727796554565, | |
| "12.up_proj": 0.47766903042793274, | |
| "12.down_proj": 0.5123953223228455, | |
| "13.q_proj": 0.38796108961105347, | |
| "13.k_proj": 0.3449362516403198, | |
| "13.v_proj": 0.5052615404129028, | |
| "13.o_proj": 0.48829060792922974, | |
| "13.gate_proj": 0.4554385840892792, | |
| "13.up_proj": 0.47726166248321533, | |
| "13.down_proj": 0.5091195106506348, | |
| "14.q_proj": 0.36075472831726074, | |
| "14.k_proj": 0.32303425669670105, | |
| "14.v_proj": 0.5109939575195312, | |
| "14.o_proj": 0.49242928624153137, | |
| "14.gate_proj": 0.4523519277572632, | |
| "14.up_proj": 0.48172876238822937, | |
| "14.down_proj": 0.532478928565979, | |
| "15.q_proj": 0.35985133051872253, | |
| "15.k_proj": 0.30060359835624695, | |
| "15.v_proj": 0.5368859767913818, | |
| "15.o_proj": 0.49520623683929443, | |
| "15.gate_proj": 0.44976216554641724, | |
| "15.up_proj": 0.47884488105773926, | |
| "15.down_proj": 0.527091383934021, | |
| "16.q_proj": 0.3671673536300659, | |
| "16.k_proj": 0.319665789604187, | |
| "16.v_proj": 0.5345363020896912, | |
| "16.o_proj": 0.48225852847099304, | |
| "16.gate_proj": 0.4529358446598053, | |
| "16.up_proj": 0.486669659614563, | |
| "16.down_proj": 0.5261300206184387, | |
| "17.q_proj": 0.39087575674057007, | |
| "17.k_proj": 0.3505445420742035, | |
| "17.v_proj": 0.5095366835594177, | |
| "17.o_proj": 0.4995217025279999, | |
| "17.gate_proj": 0.46473342180252075, | |
| "17.up_proj": 0.496961385011673, | |
| "17.down_proj": 0.5169269442558289, | |
| "18.q_proj": 0.3686104118824005, | |
| "18.k_proj": 0.3233947157859802, | |
| "18.v_proj": 0.5597348213195801, | |
| "18.o_proj": 0.47679799795150757, | |
| "18.gate_proj": 0.4632626473903656, | |
| "18.up_proj": 0.49456414580345154, | |
| "18.down_proj": 0.5329453349113464, | |
| "19.q_proj": 0.4161927103996277, | |
| "19.k_proj": 0.3401549458503723, | |
| "19.v_proj": 0.5778656601905823, | |
| "19.o_proj": 0.49888941645622253, | |
| "19.gate_proj": 0.4682750403881073, | |
| "19.up_proj": 0.4958019256591797, | |
| "19.down_proj": 0.548184871673584, | |
| "20.q_proj": 0.438224732875824, | |
| "20.k_proj": 0.380808025598526, | |
| "20.v_proj": 0.5477811098098755, | |
| "20.o_proj": 0.5176671147346497, | |
| "20.gate_proj": 0.4731369912624359, | |
| "20.up_proj": 0.49586376547813416, | |
| "20.down_proj": 0.5577734112739563, | |
| "21.q_proj": 0.4684925973415375, | |
| "21.k_proj": 0.40865975618362427, | |
| "21.v_proj": 0.522392213344574, | |
| "21.o_proj": 0.5564069747924805, | |
| "21.gate_proj": 0.4738660454750061, | |
| "21.up_proj": 0.49745243787765503, | |
| "21.down_proj": 0.5537582039833069, | |
| "22.q_proj": 0.5066864490509033, | |
| "22.k_proj": 0.4351329803466797, | |
| "22.v_proj": 0.5337282419204712, | |
| "22.o_proj": 0.5765807628631592, | |
| "22.gate_proj": 0.47878557443618774, | |
| "22.up_proj": 0.4989929795265198, | |
| "22.down_proj": 0.5528429746627808, | |
| "23.q_proj": 0.49622949957847595, | |
| "23.k_proj": 0.4328460097312927, | |
| "23.v_proj": 0.5585790872573853, | |
| "23.o_proj": 0.554423451423645, | |
| "23.gate_proj": 0.48155754804611206, | |
| "23.up_proj": 0.50072181224823, | |
| "23.down_proj": 0.5464255213737488, | |
| "24.q_proj": 0.5014330148696899, | |
| "24.k_proj": 0.45270365476608276, | |
| "24.v_proj": 0.569066047668457, | |
| "24.o_proj": 0.5753594040870667, | |
| "24.gate_proj": 0.4825673997402191, | |
| "24.up_proj": 0.5023998618125916, | |
| "24.down_proj": 0.5373293161392212, | |
| "25.q_proj": 0.5077980756759644, | |
| "25.k_proj": 0.4391773045063019, | |
| "25.v_proj": 0.5462435483932495, | |
| "25.o_proj": 0.5570780634880066, | |
| "25.gate_proj": 0.4843604266643524, | |
| "25.up_proj": 0.5035067796707153, | |
| "25.down_proj": 0.5260589122772217, | |
| "26.q_proj": 0.5026710033416748, | |
| "26.k_proj": 0.4264448285102844, | |
| "26.v_proj": 0.5314569473266602, | |
| "26.o_proj": 0.537655234336853, | |
| "26.gate_proj": 0.4852832555770874, | |
| "26.up_proj": 0.5033767223358154, | |
| "26.down_proj": 0.5124109387397766, | |
| "27.q_proj": 0.5120652914047241, | |
| "27.k_proj": 0.4639483690261841, | |
| "27.v_proj": 0.5471333265304565, | |
| "27.o_proj": 0.5303576588630676, | |
| "27.gate_proj": 0.4886907935142517, | |
| "27.up_proj": 0.5048775672912598, | |
| "27.down_proj": 0.5119288563728333, | |
| "28.q_proj": 0.4723275899887085, | |
| "28.k_proj": 0.39252132177352905, | |
| "28.v_proj": 0.5004826188087463, | |
| "28.o_proj": 0.5055510997772217, | |
| "28.gate_proj": 0.4891432821750641, | |
| "28.up_proj": 0.5027062296867371, | |
| "28.down_proj": 0.5162467956542969, | |
| "29.q_proj": 0.45425698161125183, | |
| "29.k_proj": 0.39296162128448486, | |
| "29.v_proj": 0.529412031173706, | |
| "29.o_proj": 0.4873389005661011, | |
| "29.gate_proj": 0.4842919409275055, | |
| "29.up_proj": 0.4974598288536072, | |
| "29.down_proj": 0.5207532644271851, | |
| "30.q_proj": 0.4333595335483551, | |
| "30.k_proj": 0.35803964734077454, | |
| "30.v_proj": 0.4631883203983307, | |
| "30.o_proj": 0.47360387444496155, | |
| "30.gate_proj": 0.4840942919254303, | |
| "30.up_proj": 0.48931118845939636, | |
| "30.down_proj": 0.5056915283203125, | |
| "31.q_proj": 0.38473233580589294, | |
| "31.k_proj": 0.3066442310810089, | |
| "31.v_proj": 0.4090970456600189, | |
| "31.o_proj": 0.3971359431743622, | |
| "31.gate_proj": 0.4917009174823761, | |
| "31.up_proj": 0.4686800241470337, | |
| "31.down_proj": 0.4488861858844757 | |
| }, | |
| "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": 0.7003361628605769, | |
| "mean_rel_err": 0.46043376145618303, | |
| "max_rel_err": 0.5793784856796265 | |
| }, | |
| "factorisation": { | |
| "max_rel_err": 0.00140141521114856, | |
| "factor_params": 4887871488, | |
| "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)" | |
| } | |
| } |