Text Generation
Transformers
Safetensors
mistral
svd
compression
safety
interpretability
conversational
text-generation-inference
Instructions to use Jeesup/svd-safety-mis7_swift_remove20 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Jeesup/svd-safety-mis7_swift_remove20 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Jeesup/svd-safety-mis7_swift_remove20") 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_remove20") model = AutoModelForCausalLM.from_pretrained("Jeesup/svd-safety-mis7_swift_remove20", 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_remove20 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_remove20" # 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_remove20", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove20
- SGLang
How to use Jeesup/svd-safety-mis7_swift_remove20 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_remove20" \ --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_remove20", "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_remove20" \ --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_remove20", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Jeesup/svd-safety-mis7_swift_remove20 with Docker Model Runner:
docker model run hf.co/Jeesup/svd-safety-mis7_swift_remove20
Download compression.json from Jeesup/svd-safety-mis7_swift_remove20: direct link, hf CLI and curl.
- Browser
- Download file 15.3 kB
-
https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove20/resolve/main/compression.json
- Command line
-
hf download hf://Jeesup/svd-safety-mis7_swift_remove20/compression.json
-
curl -L -o compression.json https://huggingface.co/Jeesup/svd-safety-mis7_swift_remove20/resolve/main/compression.json
15.3 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": 0.2, | |
| "internal_ratio_kept": 0.8, | |
| "from_pt": "/workspace/svd_safety/results/swift_ft/cells/mis7_swift_remove20/finetune_v/merge.pt", | |
| "mean_rel_err": 0.40079765720292926, | |
| "max_rel_err": 0.5248842835426331, | |
| "param_fraction": 0.8004338191105769, | |
| "ranks": { | |
| "0.q_proj": 1843, | |
| "0.k_proj": 776, | |
| "0.v_proj": 706, | |
| "0.o_proj": 1683, | |
| "0.gate_proj": 2688, | |
| "0.up_proj": 2688, | |
| "0.down_proj": 2723, | |
| "1.q_proj": 1851, | |
| "1.k_proj": 789, | |
| "1.v_proj": 743, | |
| "1.o_proj": 1733, | |
| "1.gate_proj": 2688, | |
| "1.up_proj": 2688, | |
| "1.down_proj": 3161, | |
| "2.q_proj": 1508, | |
| "2.k_proj": 598, | |
| "2.v_proj": 587, | |
| "2.o_proj": 1451, | |
| "2.gate_proj": 2688, | |
| "2.up_proj": 2688, | |
| "2.down_proj": 2163, | |
| "3.q_proj": 1504, | |
| "3.k_proj": 595, | |
| "3.v_proj": 581, | |
| "3.o_proj": 1437, | |
| "3.gate_proj": 2688, | |
| "3.up_proj": 2688, | |
| "3.down_proj": 2166, | |
| "4.q_proj": 1501, | |
| "4.k_proj": 594, | |
| "4.v_proj": 586, | |
| "4.o_proj": 1445, | |
| "4.gate_proj": 2688, | |
| "4.up_proj": 2688, | |
| "4.down_proj": 2193, | |
| "5.q_proj": 1501, | |
| "5.k_proj": 595, | |
| "5.v_proj": 586, | |
| "5.o_proj": 1456, | |
| "5.gate_proj": 2688, | |
| "5.up_proj": 2688, | |
| "5.down_proj": 2216, | |
| "6.q_proj": 1503, | |
| "6.k_proj": 596, | |
| "6.v_proj": 588, | |
| "6.o_proj": 1468, | |
| "6.gate_proj": 2688, | |
| "6.up_proj": 2688, | |
| "6.down_proj": 2233, | |
| "7.q_proj": 1501, | |
| "7.k_proj": 597, | |
| "7.v_proj": 589, | |
| "7.o_proj": 1480, | |
| "7.gate_proj": 2688, | |
| "7.up_proj": 2688, | |
| "7.down_proj": 2250, | |
| "8.q_proj": 1497, | |
| "8.k_proj": 593, | |
| "8.v_proj": 589, | |
| "8.o_proj": 1479, | |
| "8.gate_proj": 2688, | |
| "8.up_proj": 2688, | |
| "8.down_proj": 2246, | |
| "9.q_proj": 1494, | |
| "9.k_proj": 592, | |
| "9.v_proj": 590, | |
| "9.o_proj": 1479, | |
| "9.gate_proj": 2688, | |
| "9.up_proj": 2688, | |
| "9.down_proj": 2255, | |
| "10.q_proj": 1496, | |
| "10.k_proj": 593, | |
| "10.v_proj": 590, | |
| "10.o_proj": 1491, | |
| "10.gate_proj": 2688, | |
| "10.up_proj": 2688, | |
| "10.down_proj": 2256, | |
| "11.q_proj": 1494, | |
| "11.k_proj": 592, | |
| "11.v_proj": 592, | |
| "11.o_proj": 1495, | |
| "11.gate_proj": 2688, | |
| "11.up_proj": 2688, | |
| "11.down_proj": 2260, | |
| "12.q_proj": 1503, | |
| "12.k_proj": 598, | |
| "12.v_proj": 597, | |
| "12.o_proj": 1506, | |
| "12.gate_proj": 2688, | |
| "12.up_proj": 2688, | |
| "12.down_proj": 2285, | |
| "13.q_proj": 1498, | |
| "13.k_proj": 596, | |
| "13.v_proj": 595, | |
| "13.o_proj": 1503, | |
| "13.gate_proj": 2688, | |
| "13.up_proj": 2688, | |
| "13.down_proj": 2285, | |
| "14.q_proj": 1510, | |
| "14.k_proj": 601, | |
| "14.v_proj": 604, | |
| "14.o_proj": 1519, | |
| "14.gate_proj": 2688, | |
| "14.up_proj": 2688, | |
| "14.down_proj": 2323, | |
| "15.q_proj": 1527, | |
| "15.k_proj": 612, | |
| "15.v_proj": 615, | |
| "15.o_proj": 1542, | |
| "15.gate_proj": 2688, | |
| "15.up_proj": 2688, | |
| "15.down_proj": 2369, | |
| "16.q_proj": 1534, | |
| "16.k_proj": 615, | |
| "16.v_proj": 618, | |
| "16.o_proj": 1549, | |
| "16.gate_proj": 2688, | |
| "16.up_proj": 2688, | |
| "16.down_proj": 2399, | |
| "17.q_proj": 1565, | |
| "17.k_proj": 631, | |
| "17.v_proj": 635, | |
| "17.o_proj": 1584, | |
| "17.gate_proj": 2688, | |
| "17.up_proj": 2688, | |
| "17.down_proj": 2469, | |
| "18.q_proj": 1565, | |
| "18.k_proj": 630, | |
| "18.v_proj": 636, | |
| "18.o_proj": 1588, | |
| "18.gate_proj": 2688, | |
| "18.up_proj": 2688, | |
| "18.down_proj": 2487, | |
| "19.q_proj": 1544, | |
| "19.k_proj": 620, | |
| "19.v_proj": 626, | |
| "19.o_proj": 1572, | |
| "19.gate_proj": 2688, | |
| "19.up_proj": 2688, | |
| "19.down_proj": 2458, | |
| "20.q_proj": 1507, | |
| "20.k_proj": 599, | |
| "20.v_proj": 606, | |
| "20.o_proj": 1527, | |
| "20.gate_proj": 2688, | |
| "20.up_proj": 2688, | |
| "20.down_proj": 2381, | |
| "21.q_proj": 1497, | |
| "21.k_proj": 593, | |
| "21.v_proj": 601, | |
| "21.o_proj": 1515, | |
| "21.gate_proj": 2688, | |
| "21.up_proj": 2688, | |
| "21.down_proj": 2363, | |
| "22.q_proj": 1482, | |
| "22.k_proj": 583, | |
| "22.v_proj": 594, | |
| "22.o_proj": 1495, | |
| "22.gate_proj": 2688, | |
| "22.up_proj": 2688, | |
| "22.down_proj": 2332, | |
| "23.q_proj": 1483, | |
| "23.k_proj": 584, | |
| "23.v_proj": 593, | |
| "23.o_proj": 1505, | |
| "23.gate_proj": 2688, | |
| "23.up_proj": 2688, | |
| "23.down_proj": 2336, | |
| "24.q_proj": 1482, | |
| "24.k_proj": 584, | |
| "24.v_proj": 594, | |
| "24.o_proj": 1511, | |
| "24.gate_proj": 2688, | |
| "24.up_proj": 2688, | |
| "24.down_proj": 2338, | |
| "25.q_proj": 1486, | |
| "25.k_proj": 586, | |
| "25.v_proj": 597, | |
| "25.o_proj": 1521, | |
| "25.gate_proj": 2688, | |
| "25.up_proj": 2688, | |
| "25.down_proj": 2354, | |
| "26.q_proj": 1493, | |
| "26.k_proj": 589, | |
| "26.v_proj": 602, | |
| "26.o_proj": 1532, | |
| "26.gate_proj": 2688, | |
| "26.up_proj": 2688, | |
| "26.down_proj": 2373, | |
| "27.q_proj": 1498, | |
| "27.k_proj": 592, | |
| "27.v_proj": 606, | |
| "27.o_proj": 1532, | |
| "27.gate_proj": 2688, | |
| "27.up_proj": 2688, | |
| "27.down_proj": 2389, | |
| "28.q_proj": 1518, | |
| "28.k_proj": 603, | |
| "28.v_proj": 618, | |
| "28.o_proj": 1561, | |
| "28.gate_proj": 2688, | |
| "28.up_proj": 2688, | |
| "28.down_proj": 2451, | |
| "29.q_proj": 1530, | |
| "29.k_proj": 609, | |
| "29.v_proj": 626, | |
| "29.o_proj": 1591, | |
| "29.gate_proj": 2688, | |
| "29.up_proj": 2688, | |
| "29.down_proj": 2480, | |
| "30.q_proj": 1553, | |
| "30.k_proj": 621, | |
| "30.v_proj": 642, | |
| "30.o_proj": 1620, | |
| "30.gate_proj": 2688, | |
| "30.up_proj": 2688, | |
| "30.down_proj": 2549, | |
| "31.q_proj": 1684, | |
| "31.k_proj": 692, | |
| "31.v_proj": 716, | |
| "31.o_proj": 1782, | |
| "31.gate_proj": 2688, | |
| "31.up_proj": 2688, | |
| "31.down_proj": 2905 | |
| }, | |
| "rel_err": { | |
| "0.q_proj": 0.08500257134437561, | |
| "0.k_proj": 0.0716991052031517, | |
| "0.v_proj": 0.22834858298301697, | |
| "0.o_proj": 0.3019510805606842, | |
| "0.gate_proj": 0.26146504282951355, | |
| "0.up_proj": 0.2859381437301636, | |
| "0.down_proj": 0.3900061547756195, | |
| "1.q_proj": 0.13657206296920776, | |
| "1.k_proj": 0.07040991634130478, | |
| "1.v_proj": 0.34303975105285645, | |
| "1.o_proj": 0.4063681960105896, | |
| "1.gate_proj": 0.3133699595928192, | |
| "1.up_proj": 0.33793723583221436, | |
| "1.down_proj": 0.3351709246635437, | |
| "2.q_proj": 0.35540199279785156, | |
| "2.k_proj": 0.3083972930908203, | |
| "2.v_proj": 0.4814577102661133, | |
| "2.o_proj": 0.47539830207824707, | |
| "2.gate_proj": 0.3777889311313629, | |
| "2.up_proj": 0.39566245675086975, | |
| "2.down_proj": 0.5091328620910645, | |
| "3.q_proj": 0.3229297995567322, | |
| "3.k_proj": 0.25059062242507935, | |
| "3.v_proj": 0.5148245096206665, | |
| "3.o_proj": 0.5196603536605835, | |
| "3.gate_proj": 0.38781943917274475, | |
| "3.up_proj": 0.4100722670555115, | |
| "3.down_proj": 0.49922603368759155, | |
| "4.q_proj": 0.3310073912143707, | |
| "4.k_proj": 0.2755686044692993, | |
| "4.v_proj": 0.4732750952243805, | |
| "4.o_proj": 0.4666188955307007, | |
| "4.gate_proj": 0.3899378180503845, | |
| "4.up_proj": 0.4183727204799652, | |
| "4.down_proj": 0.5054365992546082, | |
| "5.q_proj": 0.3576917052268982, | |
| "5.k_proj": 0.3083855211734772, | |
| "5.v_proj": 0.5068833231925964, | |
| "5.o_proj": 0.496389240026474, | |
| "5.gate_proj": 0.3855798542499542, | |
| "5.up_proj": 0.4201187789440155, | |
| "5.down_proj": 0.4902642071247101, | |
| "6.q_proj": 0.346874475479126, | |
| "6.k_proj": 0.2829170227050781, | |
| "6.v_proj": 0.4889771640300751, | |
| "6.o_proj": 0.48441195487976074, | |
| "6.gate_proj": 0.39520981907844543, | |
| "6.up_proj": 0.4232458472251892, | |
| "6.down_proj": 0.4810233414173126, | |
| "7.q_proj": 0.33142852783203125, | |
| "7.k_proj": 0.2655273973941803, | |
| "7.v_proj": 0.4808761775493622, | |
| "7.o_proj": 0.4556846618652344, | |
| "7.gate_proj": 0.387683242559433, | |
| "7.up_proj": 0.4178556799888611, | |
| "7.down_proj": 0.4722205698490143, | |
| "8.q_proj": 0.33769723773002625, | |
| "8.k_proj": 0.2662927508354187, | |
| "8.v_proj": 0.4425075054168701, | |
| "8.o_proj": 0.44497787952423096, | |
| "8.gate_proj": 0.3896368443965912, | |
| "8.up_proj": 0.41572481393814087, | |
| "8.down_proj": 0.4673706591129303, | |
| "9.q_proj": 0.3340452313423157, | |
| "9.k_proj": 0.2718675434589386, | |
| "9.v_proj": 0.4268815815448761, | |
| "9.o_proj": 0.44764506816864014, | |
| "9.gate_proj": 0.38748565316200256, | |
| "9.up_proj": 0.4152057468891144, | |
| "9.down_proj": 0.4528351426124573, | |
| "10.q_proj": 0.3335397243499756, | |
| "10.k_proj": 0.2776794731616974, | |
| "10.v_proj": 0.4279860556125641, | |
| "10.o_proj": 0.44602537155151367, | |
| "10.gate_proj": 0.38787901401519775, | |
| "10.up_proj": 0.4141860604286194, | |
| "10.down_proj": 0.46012479066848755, | |
| "11.q_proj": 0.3297421634197235, | |
| "11.k_proj": 0.26538947224617004, | |
| "11.v_proj": 0.3986572325229645, | |
| "11.o_proj": 0.4476287066936493, | |
| "11.gate_proj": 0.38680461049079895, | |
| "11.up_proj": 0.41310203075408936, | |
| "11.down_proj": 0.45812714099884033, | |
| "12.q_proj": 0.319181889295578, | |
| "12.k_proj": 0.24617546796798706, | |
| "12.v_proj": 0.4383234977722168, | |
| "12.o_proj": 0.42985114455223083, | |
| "12.gate_proj": 0.3842836320400238, | |
| "12.up_proj": 0.4086284637451172, | |
| "12.down_proj": 0.45287853479385376, | |
| "13.q_proj": 0.342117041349411, | |
| "13.k_proj": 0.28727155923843384, | |
| "13.v_proj": 0.4403918981552124, | |
| "13.o_proj": 0.43300414085388184, | |
| "13.gate_proj": 0.3891292214393616, | |
| "13.up_proj": 0.4081304669380188, | |
| "13.down_proj": 0.4504336416721344, | |
| "14.q_proj": 0.31687843799591064, | |
| "14.k_proj": 0.26946091651916504, | |
| "14.v_proj": 0.44405218958854675, | |
| "14.o_proj": 0.43850401043891907, | |
| "14.gate_proj": 0.38567572832107544, | |
| "14.up_proj": 0.4120667278766632, | |
| "14.down_proj": 0.4714415967464447, | |
| "15.q_proj": 0.31565016508102417, | |
| "15.k_proj": 0.2453945130109787, | |
| "15.v_proj": 0.4641188681125641, | |
| "15.o_proj": 0.4399622678756714, | |
| "15.gate_proj": 0.38257908821105957, | |
| "15.up_proj": 0.408530592918396, | |
| "15.down_proj": 0.4638931155204773, | |
| "16.q_proj": 0.32283511757850647, | |
| "16.k_proj": 0.2652204930782318, | |
| "16.v_proj": 0.464958131313324, | |
| "16.o_proj": 0.42660167813301086, | |
| "16.gate_proj": 0.3851650655269623, | |
| "16.up_proj": 0.41526761651039124, | |
| "16.down_proj": 0.46140268445014954, | |
| "17.q_proj": 0.3427833616733551, | |
| "17.k_proj": 0.2890065610408783, | |
| "17.v_proj": 0.4380885064601898, | |
| "17.o_proj": 0.44451603293418884, | |
| "17.gate_proj": 0.3959171772003174, | |
| "17.up_proj": 0.4251500368118286, | |
| "17.down_proj": 0.4505082368850708, | |
| "18.q_proj": 0.3227398097515106, | |
| "18.k_proj": 0.2645529508590698, | |
| "18.v_proj": 0.486581027507782, | |
| "18.o_proj": 0.4234537184238434, | |
| "18.gate_proj": 0.39415696263313293, | |
| "18.up_proj": 0.42284369468688965, | |
| "18.down_proj": 0.4652669429779053, | |
| "19.q_proj": 0.36690786480903625, | |
| "19.k_proj": 0.27805012464523315, | |
| "19.v_proj": 0.5099502205848694, | |
| "19.o_proj": 0.4442979693412781, | |
| "19.gate_proj": 0.39896416664123535, | |
| "19.up_proj": 0.42378658056259155, | |
| "19.down_proj": 0.48121050000190735, | |
| "20.q_proj": 0.38949716091156006, | |
| "20.k_proj": 0.32147786021232605, | |
| "20.v_proj": 0.4748760461807251, | |
| "20.o_proj": 0.46483227610588074, | |
| "20.gate_proj": 0.4035407602787018, | |
| "20.up_proj": 0.423808217048645, | |
| "20.down_proj": 0.49269551038742065, | |
| "21.q_proj": 0.41730770468711853, | |
| "21.k_proj": 0.3455001711845398, | |
| "21.v_proj": 0.45047393441200256, | |
| "21.o_proj": 0.5031458139419556, | |
| "21.gate_proj": 0.4041469395160675, | |
| "21.up_proj": 0.4253368079662323, | |
| "21.down_proj": 0.4889674484729767, | |
| "22.q_proj": 0.4534510374069214, | |
| "22.k_proj": 0.37348392605781555, | |
| "22.v_proj": 0.4685239791870117, | |
| "22.o_proj": 0.5248842835426331, | |
| "22.gate_proj": 0.4087425768375397, | |
| "22.up_proj": 0.42668619751930237, | |
| "22.down_proj": 0.4892253875732422, | |
| "23.q_proj": 0.4440714418888092, | |
| "23.k_proj": 0.3682272434234619, | |
| "23.v_proj": 0.49346232414245605, | |
| "23.o_proj": 0.502592921257019, | |
| "23.gate_proj": 0.4109232425689697, | |
| "23.up_proj": 0.42807736992836, | |
| "23.down_proj": 0.48296502232551575, | |
| "24.q_proj": 0.4487209916114807, | |
| "24.k_proj": 0.38935786485671997, | |
| "24.v_proj": 0.5036438703536987, | |
| "24.o_proj": 0.5222789645195007, | |
| "24.gate_proj": 0.4119880795478821, | |
| "24.up_proj": 0.42954087257385254, | |
| "24.down_proj": 0.47408875823020935, | |
| "25.q_proj": 0.4551296532154083, | |
| "25.k_proj": 0.37871479988098145, | |
| "25.v_proj": 0.4806521534919739, | |
| "25.o_proj": 0.5046061873435974, | |
| "25.gate_proj": 0.4138759672641754, | |
| "25.up_proj": 0.4307102560997009, | |
| "25.down_proj": 0.463013231754303, | |
| "26.q_proj": 0.44990450143814087, | |
| "26.k_proj": 0.3649565577507019, | |
| "26.v_proj": 0.46789830923080444, | |
| "26.o_proj": 0.4847952425479889, | |
| "26.gate_proj": 0.414262056350708, | |
| "26.up_proj": 0.43036970496177673, | |
| "26.down_proj": 0.4495799243450165, | |
| "27.q_proj": 0.4583161473274231, | |
| "27.k_proj": 0.39734572172164917, | |
| "27.v_proj": 0.4797323942184448, | |
| "27.o_proj": 0.47943976521492004, | |
| "27.gate_proj": 0.41736742854118347, | |
| "27.up_proj": 0.4318309724330902, | |
| "27.down_proj": 0.44924622774124146, | |
| "28.q_proj": 0.41977617144584656, | |
| "28.k_proj": 0.3314531147480011, | |
| "28.v_proj": 0.4329354465007782, | |
| "28.o_proj": 0.4519560635089874, | |
| "28.gate_proj": 0.41815441846847534, | |
| "28.up_proj": 0.4296439588069916, | |
| "28.down_proj": 0.4507562816143036, | |
| "29.q_proj": 0.4013705253601074, | |
| "29.k_proj": 0.3302054703235626, | |
| "29.v_proj": 0.45994922518730164, | |
| "29.o_proj": 0.4338970482349396, | |
| "29.gate_proj": 0.4139324426651001, | |
| "29.up_proj": 0.4252754747867584, | |
| "29.down_proj": 0.4538421928882599, | |
| "30.q_proj": 0.3819069266319275, | |
| "30.k_proj": 0.300334095954895, | |
| "30.v_proj": 0.3898763358592987, | |
| "30.o_proj": 0.4189046621322632, | |
| "30.gate_proj": 0.4136664569377899, | |
| "30.up_proj": 0.4180448055267334, | |
| "30.down_proj": 0.43752264976501465, | |
| "31.q_proj": 0.3336416184902191, | |
| "31.k_proj": 0.2430993765592575, | |
| "31.v_proj": 0.32917752861976624, | |
| "31.o_proj": 0.3434261381626129, | |
| "31.gate_proj": 0.4216300845146179, | |
| "31.up_proj": 0.40061134099960327, | |
| "31.down_proj": 0.3710506558418274 | |
| }, | |
| "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": 0.8004338191105769, | |
| "mean_rel_err": 0.3994709092990628, | |
| "max_rel_err": 0.5239740610122681 | |
| }, | |
| "factorisation": { | |
| "max_rel_err": 0.0013197725638747215, | |
| "factor_params": 5586485248, | |
| "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)" | |
| } | |
| } |