Text Generation
Transformers
Safetensors
llama
llama2
svd
compression
safety
interpretability
conversational
text-generation-inference
Instructions to use Jeesup/svd-safety-l2_basis_remove30 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Jeesup/svd-safety-l2_basis_remove30 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Jeesup/svd-safety-l2_basis_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-l2_basis_remove30") model = AutoModelForCausalLM.from_pretrained("Jeesup/svd-safety-l2_basis_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-l2_basis_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-l2_basis_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-l2_basis_remove30", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Jeesup/svd-safety-l2_basis_remove30
- SGLang
How to use Jeesup/svd-safety-l2_basis_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-l2_basis_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-l2_basis_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-l2_basis_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-l2_basis_remove30", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Jeesup/svd-safety-l2_basis_remove30 with Docker Model Runner:
docker model run hf.co/Jeesup/svd-safety-l2_basis_remove30
Download compression.json from Jeesup/svd-safety-l2_basis_remove30: direct link, hf CLI and curl.
- Browser
- Download file 15.9 kB
-
https://huggingface.co/Jeesup/svd-safety-l2_basis_remove30/resolve/main/compression.json
- Command line
-
hf download hf://Jeesup/svd-safety-l2_basis_remove30/compression.json
-
curl -L -o compression.json https://huggingface.co/Jeesup/svd-safety-l2_basis_remove30/resolve/main/compression.json
15.9 kB
| { | |
| "source": "/workspace/svd_safety/assets/models/Llama-2-7b-chat-hf", | |
| "upstream": "TUDa-HWAI/Basis_Sharing", | |
| "step": "basis_sharing", | |
| "ratio_removed": 0.3, | |
| "group_size": 2, | |
| "nsamples": 256, | |
| "seqlen": 2048, | |
| "seed": 42, | |
| "share_part": [ | |
| "v", | |
| "k", | |
| "q", | |
| "up", | |
| "gate" | |
| ], | |
| "private_part": [ | |
| "down", | |
| "o" | |
| ], | |
| "factored": "/workspace/svd_safety/results/basis_sharing/cells/l2_basis_remove30/truncated/factored.pt", | |
| "param_fraction": 0.6998272040965026, | |
| "ranks": { | |
| "0.q_proj": 1911, | |
| "0.k_proj": 1911, | |
| "0.v_proj": 1911, | |
| "0.o_proj": 1433, | |
| "0.gate_proj": 2417, | |
| "0.up_proj": 2417, | |
| "0.down_proj": 2089, | |
| "1.q_proj": 1911, | |
| "1.k_proj": 1911, | |
| "1.v_proj": 1911, | |
| "1.o_proj": 1433, | |
| "1.gate_proj": 2417, | |
| "1.up_proj": 2417, | |
| "1.down_proj": 2089, | |
| "2.q_proj": 1911, | |
| "2.k_proj": 1911, | |
| "2.v_proj": 1911, | |
| "2.o_proj": 1433, | |
| "2.gate_proj": 2417, | |
| "2.up_proj": 2417, | |
| "2.down_proj": 2089, | |
| "3.q_proj": 1911, | |
| "3.k_proj": 1911, | |
| "3.v_proj": 1911, | |
| "3.o_proj": 1433, | |
| "3.gate_proj": 2417, | |
| "3.up_proj": 2417, | |
| "3.down_proj": 2089, | |
| "4.q_proj": 1911, | |
| "4.k_proj": 1911, | |
| "4.v_proj": 1911, | |
| "4.o_proj": 1433, | |
| "4.gate_proj": 2417, | |
| "4.up_proj": 2417, | |
| "4.down_proj": 2089, | |
| "5.q_proj": 1911, | |
| "5.k_proj": 1911, | |
| "5.v_proj": 1911, | |
| "5.o_proj": 1433, | |
| "5.gate_proj": 2417, | |
| "5.up_proj": 2417, | |
| "5.down_proj": 2089, | |
| "6.q_proj": 1911, | |
| "6.k_proj": 1911, | |
| "6.v_proj": 1911, | |
| "6.o_proj": 1433, | |
| "6.gate_proj": 2417, | |
| "6.up_proj": 2417, | |
| "6.down_proj": 2089, | |
| "7.q_proj": 1911, | |
| "7.k_proj": 1911, | |
| "7.v_proj": 1911, | |
| "7.o_proj": 1433, | |
| "7.gate_proj": 2417, | |
| "7.up_proj": 2417, | |
| "7.down_proj": 2089, | |
| "8.q_proj": 1911, | |
| "8.k_proj": 1911, | |
| "8.v_proj": 1911, | |
| "8.o_proj": 1433, | |
| "8.gate_proj": 2417, | |
| "8.up_proj": 2417, | |
| "8.down_proj": 2089, | |
| "9.q_proj": 1911, | |
| "9.k_proj": 1911, | |
| "9.v_proj": 1911, | |
| "9.o_proj": 1433, | |
| "9.gate_proj": 2417, | |
| "9.up_proj": 2417, | |
| "9.down_proj": 2089, | |
| "10.q_proj": 1911, | |
| "10.k_proj": 1911, | |
| "10.v_proj": 1911, | |
| "10.o_proj": 1433, | |
| "10.gate_proj": 2417, | |
| "10.up_proj": 2417, | |
| "10.down_proj": 2089, | |
| "11.q_proj": 1911, | |
| "11.k_proj": 1911, | |
| "11.v_proj": 1911, | |
| "11.o_proj": 1433, | |
| "11.gate_proj": 2417, | |
| "11.up_proj": 2417, | |
| "11.down_proj": 2089, | |
| "12.q_proj": 1911, | |
| "12.k_proj": 1911, | |
| "12.v_proj": 1911, | |
| "12.o_proj": 1433, | |
| "12.gate_proj": 2417, | |
| "12.up_proj": 2417, | |
| "12.down_proj": 2089, | |
| "13.q_proj": 1911, | |
| "13.k_proj": 1911, | |
| "13.v_proj": 1911, | |
| "13.o_proj": 1433, | |
| "13.gate_proj": 2417, | |
| "13.up_proj": 2417, | |
| "13.down_proj": 2089, | |
| "14.q_proj": 1911, | |
| "14.k_proj": 1911, | |
| "14.v_proj": 1911, | |
| "14.o_proj": 1433, | |
| "14.gate_proj": 2417, | |
| "14.up_proj": 2417, | |
| "14.down_proj": 2089, | |
| "15.q_proj": 1911, | |
| "15.k_proj": 1911, | |
| "15.v_proj": 1911, | |
| "15.o_proj": 1433, | |
| "15.gate_proj": 2417, | |
| "15.up_proj": 2417, | |
| "15.down_proj": 2089, | |
| "16.q_proj": 1911, | |
| "16.k_proj": 1911, | |
| "16.v_proj": 1911, | |
| "16.o_proj": 1433, | |
| "16.gate_proj": 2417, | |
| "16.up_proj": 2417, | |
| "16.down_proj": 2089, | |
| "17.q_proj": 1911, | |
| "17.k_proj": 1911, | |
| "17.v_proj": 1911, | |
| "17.o_proj": 1433, | |
| "17.gate_proj": 2417, | |
| "17.up_proj": 2417, | |
| "17.down_proj": 2089, | |
| "18.q_proj": 1911, | |
| "18.k_proj": 1911, | |
| "18.v_proj": 1911, | |
| "18.o_proj": 1433, | |
| "18.gate_proj": 2417, | |
| "18.up_proj": 2417, | |
| "18.down_proj": 2089, | |
| "19.q_proj": 1911, | |
| "19.k_proj": 1911, | |
| "19.v_proj": 1911, | |
| "19.o_proj": 1433, | |
| "19.gate_proj": 2417, | |
| "19.up_proj": 2417, | |
| "19.down_proj": 2089, | |
| "20.q_proj": 1911, | |
| "20.k_proj": 1911, | |
| "20.v_proj": 1911, | |
| "20.o_proj": 1433, | |
| "20.gate_proj": 2417, | |
| "20.up_proj": 2417, | |
| "20.down_proj": 2089, | |
| "21.q_proj": 1911, | |
| "21.k_proj": 1911, | |
| "21.v_proj": 1911, | |
| "21.o_proj": 1433, | |
| "21.gate_proj": 2417, | |
| "21.up_proj": 2417, | |
| "21.down_proj": 2089, | |
| "22.q_proj": 1911, | |
| "22.k_proj": 1911, | |
| "22.v_proj": 1911, | |
| "22.o_proj": 1433, | |
| "22.gate_proj": 2417, | |
| "22.up_proj": 2417, | |
| "22.down_proj": 2089, | |
| "23.q_proj": 1911, | |
| "23.k_proj": 1911, | |
| "23.v_proj": 1911, | |
| "23.o_proj": 1433, | |
| "23.gate_proj": 2417, | |
| "23.up_proj": 2417, | |
| "23.down_proj": 2089, | |
| "24.q_proj": 1911, | |
| "24.k_proj": 1911, | |
| "24.v_proj": 1911, | |
| "24.o_proj": 1433, | |
| "24.gate_proj": 2417, | |
| "24.up_proj": 2417, | |
| "24.down_proj": 2089, | |
| "25.q_proj": 1911, | |
| "25.k_proj": 1911, | |
| "25.v_proj": 1911, | |
| "25.o_proj": 1433, | |
| "25.gate_proj": 2417, | |
| "25.up_proj": 2417, | |
| "25.down_proj": 2089, | |
| "26.q_proj": 1911, | |
| "26.k_proj": 1911, | |
| "26.v_proj": 1911, | |
| "26.o_proj": 1433, | |
| "26.gate_proj": 2417, | |
| "26.up_proj": 2417, | |
| "26.down_proj": 2089, | |
| "27.q_proj": 1911, | |
| "27.k_proj": 1911, | |
| "27.v_proj": 1911, | |
| "27.o_proj": 1433, | |
| "27.gate_proj": 2417, | |
| "27.up_proj": 2417, | |
| "27.down_proj": 2089, | |
| "28.q_proj": 1911, | |
| "28.k_proj": 1911, | |
| "28.v_proj": 1911, | |
| "28.o_proj": 1433, | |
| "28.gate_proj": 2417, | |
| "28.up_proj": 2417, | |
| "28.down_proj": 2089, | |
| "29.q_proj": 1911, | |
| "29.k_proj": 1911, | |
| "29.v_proj": 1911, | |
| "29.o_proj": 1433, | |
| "29.gate_proj": 2417, | |
| "29.up_proj": 2417, | |
| "29.down_proj": 2089, | |
| "30.q_proj": 1911, | |
| "30.k_proj": 1911, | |
| "30.v_proj": 1911, | |
| "30.o_proj": 1433, | |
| "30.gate_proj": 2417, | |
| "30.up_proj": 2417, | |
| "30.down_proj": 2089, | |
| "31.q_proj": 1911, | |
| "31.k_proj": 1911, | |
| "31.v_proj": 1911, | |
| "31.o_proj": 1433, | |
| "31.gate_proj": 2417, | |
| "31.up_proj": 2417, | |
| "31.down_proj": 2089 | |
| }, | |
| "stage": "basis_recovery_finetuned", | |
| "recovery_finetune": { | |
| "protocol": "basis_sharing_coefficient_lora", | |
| "script": "gitclones/SVD-LLM/utils/LoRA.py", | |
| "data_path": "alpaca-cleaned (/workspace/svd_safety/data/alpaca_cleaned.json)", | |
| "epochs": 2, | |
| "learning_rate": 0.0001, | |
| "batch_size": 64, | |
| "micro_batch_size": 4, | |
| "cutoff_len": 256, | |
| "val_set_size": 2000, | |
| "max_train_samples": "all", | |
| "precision": "bf16 (upstream hard-codes fp16; see run_with_torch_compat)", | |
| "is_methods_own_stage2": false, | |
| "note": "Basis Sharing publishes its own LoRA (lora.py: wikitext, batch 1, peft default q/v coefficient targets). This cell uses the project's SVD-LLM alpaca recipe so recovery data is constant across compressors, on all seven coefficient types and never on a basis.", | |
| "lora_r": 8, | |
| "lora_alpha": 16, | |
| "lora_dropout": 0.05, | |
| "lora_targets": [ | |
| "down_proj", | |
| "gate_proj", | |
| "k_proj", | |
| "o_proj", | |
| "q_proj", | |
| "up_proj", | |
| "v_proj" | |
| ], | |
| "trained": "Coefficient modules only; shared and private bases frozen", | |
| "merge": "C' = C + (alpha/r) B_lora A_lora, then W = C' @ basis", | |
| "mean_rel_delta_coefficient": 0.026299988264716894 | |
| }, | |
| "param_fraction_is_upper_bound": false, | |
| "rel_delta_coefficient": { | |
| "0.down_proj": 0.01571832410991192, | |
| "0.gate_proj": 0.07073784619569778, | |
| "0.up_proj": 0.051439255475997925, | |
| "0.k_proj": 0.03125172480940819, | |
| "0.o_proj": 0.014449120499193668, | |
| "0.q_proj": 0.03769339993596077, | |
| "0.v_proj": 0.019454704597592354, | |
| "1.down_proj": 0.007780988700687885, | |
| "1.gate_proj": 0.05042870715260506, | |
| "1.up_proj": 0.04310716688632965, | |
| "1.k_proj": 0.01711668074131012, | |
| "1.o_proj": 0.013755874708294868, | |
| "1.q_proj": 0.01987244002521038, | |
| "1.v_proj": 0.015976836904883385, | |
| "2.down_proj": 0.01568828709423542, | |
| "2.gate_proj": 0.04502059519290924, | |
| "2.up_proj": 0.04337015375494957, | |
| "2.k_proj": 0.01923982799053192, | |
| "2.o_proj": 0.018201353028416634, | |
| "2.q_proj": 0.019387632608413696, | |
| "2.v_proj": 0.025767281651496887, | |
| "3.down_proj": 0.015289364382624626, | |
| "3.gate_proj": 0.03861368075013161, | |
| "3.up_proj": 0.040375787764787674, | |
| "3.k_proj": 0.019376985728740692, | |
| "3.o_proj": 0.017843443900346756, | |
| "3.q_proj": 0.01794094406068325, | |
| "3.v_proj": 0.025516241788864136, | |
| "4.down_proj": 0.013919559307396412, | |
| "4.gate_proj": 0.03799600154161453, | |
| "4.up_proj": 0.03619149327278137, | |
| "4.k_proj": 0.021520331501960754, | |
| "4.o_proj": 0.017562542110681534, | |
| "4.q_proj": 0.019273044541478157, | |
| "4.v_proj": 0.02358822524547577, | |
| "5.down_proj": 0.014357398264110088, | |
| "5.gate_proj": 0.03674551099538803, | |
| "5.up_proj": 0.033642545342445374, | |
| "5.k_proj": 0.019489075988531113, | |
| "5.o_proj": 0.01640571840107441, | |
| "5.q_proj": 0.017913563176989555, | |
| "5.v_proj": 0.02251928113400936, | |
| "6.down_proj": 0.013552287593483925, | |
| "6.gate_proj": 0.037070877850055695, | |
| "6.up_proj": 0.03368235006928444, | |
| "6.k_proj": 0.018359050154685974, | |
| "6.o_proj": 0.01698077842593193, | |
| "6.q_proj": 0.018099360167980194, | |
| "6.v_proj": 0.024023892357945442, | |
| "7.down_proj": 0.014833086170256138, | |
| "7.gate_proj": 0.03736833110451698, | |
| "7.up_proj": 0.03326217830181122, | |
| "7.k_proj": 0.01821022666990757, | |
| "7.o_proj": 0.014988509006798267, | |
| "7.q_proj": 0.017297787591814995, | |
| "7.v_proj": 0.022407837212085724, | |
| "8.down_proj": 0.014573442749679089, | |
| "8.gate_proj": 0.03457353264093399, | |
| "8.up_proj": 0.033223461359739304, | |
| "8.k_proj": 0.01891668699681759, | |
| "8.o_proj": 0.016290225088596344, | |
| "8.q_proj": 0.018463721498847008, | |
| "8.v_proj": 0.020800800994038582, | |
| "9.down_proj": 0.013626599684357643, | |
| "9.gate_proj": 0.03514138609170914, | |
| "9.up_proj": 0.03454797342419624, | |
| "9.k_proj": 0.01954204961657524, | |
| "9.o_proj": 0.01583862490952015, | |
| "9.q_proj": 0.018615638837218285, | |
| "9.v_proj": 0.020167481154203415, | |
| "10.down_proj": 0.014242639765143394, | |
| "10.gate_proj": 0.036446455866098404, | |
| "10.up_proj": 0.03395068645477295, | |
| "10.k_proj": 0.019043728709220886, | |
| "10.o_proj": 0.016665460541844368, | |
| "10.q_proj": 0.018057456240057945, | |
| "10.v_proj": 0.019024686887860298, | |
| "11.down_proj": 0.015268338844180107, | |
| "11.gate_proj": 0.034873563796281815, | |
| "11.up_proj": 0.03515440225601196, | |
| "11.k_proj": 0.020495600998401642, | |
| "11.o_proj": 0.01653156615793705, | |
| "11.q_proj": 0.01963174156844616, | |
| "11.v_proj": 0.018966326490044594, | |
| "12.down_proj": 0.01490208599716425, | |
| "12.gate_proj": 0.036391180008649826, | |
| "12.up_proj": 0.03552509844303131, | |
| "12.k_proj": 0.020210448652505875, | |
| "12.o_proj": 0.015827499330043793, | |
| "12.q_proj": 0.01948561705648899, | |
| "12.v_proj": 0.020654533058404922, | |
| "13.down_proj": 0.01405719481408596, | |
| "13.gate_proj": 0.03551631048321724, | |
| "13.up_proj": 0.037898264825344086, | |
| "13.k_proj": 0.020987654104828835, | |
| "13.o_proj": 0.016064412891864777, | |
| "13.q_proj": 0.019807111471891403, | |
| "13.v_proj": 0.01962696947157383, | |
| "14.down_proj": 0.016009340062737465, | |
| "14.gate_proj": 0.03802134096622467, | |
| "14.up_proj": 0.03786147013306618, | |
| "14.k_proj": 0.021262750029563904, | |
| "14.o_proj": 0.017347760498523712, | |
| "14.q_proj": 0.020732464268803596, | |
| "14.v_proj": 0.02084069512784481, | |
| "15.down_proj": 0.016418438404798508, | |
| "15.gate_proj": 0.038650088012218475, | |
| "15.up_proj": 0.035394005477428436, | |
| "15.k_proj": 0.022153930738568306, | |
| "15.o_proj": 0.01559238601475954, | |
| "15.q_proj": 0.02090977504849434, | |
| "15.v_proj": 0.02013670839369297, | |
| "16.down_proj": 0.016190309077501297, | |
| "16.gate_proj": 0.040097519755363464, | |
| "16.up_proj": 0.03531859815120697, | |
| "16.k_proj": 0.024484077468514442, | |
| "16.o_proj": 0.018780086189508438, | |
| "16.q_proj": 0.02221532352268696, | |
| "16.v_proj": 0.02362828142940998, | |
| "17.down_proj": 0.017163388431072235, | |
| "17.gate_proj": 0.039196014404296875, | |
| "17.up_proj": 0.03838655725121498, | |
| "17.k_proj": 0.023485826328396797, | |
| "17.o_proj": 0.01819983869791031, | |
| "17.q_proj": 0.024406565353274345, | |
| "17.v_proj": 0.023620449006557465, | |
| "18.down_proj": 0.017160629853606224, | |
| "18.gate_proj": 0.038988612592220306, | |
| "18.up_proj": 0.04259044677019119, | |
| "18.k_proj": 0.024072891101241112, | |
| "18.o_proj": 0.019031178206205368, | |
| "18.q_proj": 0.024145307019352913, | |
| "18.v_proj": 0.02690170519053936, | |
| "19.down_proj": 0.01757754199206829, | |
| "19.gate_proj": 0.04096819460391998, | |
| "19.up_proj": 0.03828761354088783, | |
| "19.k_proj": 0.025333957746624947, | |
| "19.o_proj": 0.01932433247566223, | |
| "19.q_proj": 0.024755822494626045, | |
| "19.v_proj": 0.02619953081011772, | |
| "20.down_proj": 0.017618484795093536, | |
| "20.gate_proj": 0.04168369621038437, | |
| "20.up_proj": 0.04031258076429367, | |
| "20.k_proj": 0.025555750355124474, | |
| "20.o_proj": 0.019319145008921623, | |
| "20.q_proj": 0.026066727936267853, | |
| "20.v_proj": 0.0259451512247324, | |
| "21.down_proj": 0.018069017678499222, | |
| "21.gate_proj": 0.03980571776628494, | |
| "21.up_proj": 0.04347342625260353, | |
| "21.k_proj": 0.026453960686922073, | |
| "21.o_proj": 0.01866300217807293, | |
| "21.q_proj": 0.025375496596097946, | |
| "21.v_proj": 0.027465440332889557, | |
| "22.down_proj": 0.01788206212222576, | |
| "22.gate_proj": 0.04741738364100456, | |
| "22.up_proj": 0.04643930122256279, | |
| "22.k_proj": 0.026044832542538643, | |
| "22.o_proj": 0.018997332081198692, | |
| "22.q_proj": 0.025562603026628494, | |
| "22.v_proj": 0.029967913404107094, | |
| "23.down_proj": 0.018382783979177475, | |
| "23.gate_proj": 0.04516542702913284, | |
| "23.up_proj": 0.04480283334851265, | |
| "23.k_proj": 0.025396348908543587, | |
| "23.o_proj": 0.016814321279525757, | |
| "23.q_proj": 0.025049729272723198, | |
| "23.v_proj": 0.0272342711687088, | |
| "24.down_proj": 0.018060794100165367, | |
| "24.gate_proj": 0.045451823621988297, | |
| "24.up_proj": 0.04517007991671562, | |
| "24.k_proj": 0.024571320042014122, | |
| "24.o_proj": 0.015839651226997375, | |
| "24.q_proj": 0.02506706677377224, | |
| "24.v_proj": 0.030967289581894875, | |
| "25.down_proj": 0.017680563032627106, | |
| "25.gate_proj": 0.04438598453998566, | |
| "25.up_proj": 0.04630052670836449, | |
| "25.k_proj": 0.026787467300891876, | |
| "25.o_proj": 0.018963590264320374, | |
| "25.q_proj": 0.024902934208512306, | |
| "25.v_proj": 0.02655676379799843, | |
| "26.down_proj": 0.01728445664048195, | |
| "26.gate_proj": 0.04320848360657692, | |
| "26.up_proj": 0.04594198614358902, | |
| "26.k_proj": 0.029789509251713753, | |
| "26.o_proj": 0.01576789654791355, | |
| "26.q_proj": 0.02549917809665203, | |
| "26.v_proj": 0.027597865089774132, | |
| "27.down_proj": 0.016805997118353844, | |
| "27.gate_proj": 0.04197928309440613, | |
| "27.up_proj": 0.04444769397377968, | |
| "27.k_proj": 0.024795103818178177, | |
| "27.o_proj": 0.023305708542466164, | |
| "27.q_proj": 0.026696208864450455, | |
| "27.v_proj": 0.02590230293571949, | |
| "28.down_proj": 0.01710508018732071, | |
| "28.gate_proj": 0.04344693571329117, | |
| "28.up_proj": 0.041666075587272644, | |
| "28.k_proj": 0.02632494829595089, | |
| "28.o_proj": 0.017413025721907616, | |
| "28.q_proj": 0.023984916508197784, | |
| "28.v_proj": 0.029418673366308212, | |
| "29.down_proj": 0.01671851985156536, | |
| "29.gate_proj": 0.0447494238615036, | |
| "29.up_proj": 0.04233136773109436, | |
| "29.k_proj": 0.028628991916775703, | |
| "29.o_proj": 0.01745348423719406, | |
| "29.q_proj": 0.02534261718392372, | |
| "29.v_proj": 0.026935601606965065, | |
| "30.down_proj": 0.013234722428023815, | |
| "30.gate_proj": 0.04196500778198242, | |
| "30.up_proj": 0.0402400940656662, | |
| "30.k_proj": 0.029129525646567345, | |
| "30.o_proj": 0.018144818022847176, | |
| "30.q_proj": 0.02681029587984085, | |
| "30.v_proj": 0.019666632637381554, | |
| "31.down_proj": 0.012836103327572346, | |
| "31.gate_proj": 0.04107679799199104, | |
| "31.up_proj": 0.038529664278030396, | |
| "31.k_proj": 0.033828120678663254, | |
| "31.o_proj": 0.016145970672369003, | |
| "31.q_proj": 0.03200652077794075, | |
| "31.v_proj": 0.022225258871912956 | |
| } | |
| } |