Instructions to use IFM/K2-Horizon-7B-Uno with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use IFM/K2-Horizon-7B-Uno with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("IFM/K2-Horizon-7B") model = PeftModel.from_pretrained(base_model, "IFM/K2-Horizon-7B-Uno") - Notebooks
- Google Colab
- Kaggle
Download source_info.json from IFM/K2-Horizon-7B-Uno: direct link, hf CLI and curl.
- Browser
- Download file 2.61 kB
-
https://huggingface.co/IFM/K2-Horizon-7B-Uno/resolve/main/source_info.json
- Command line
-
hf download hf://IFM/K2-Horizon-7B-Uno/source_info.json
-
curl -L -o source_info.json https://huggingface.co/IFM/K2-Horizon-7B-Uno/resolve/main/source_info.json
2.61 kB
| { | |
| "adapter_config": "adapter_config.json", | |
| "adapter_model": "adapter_model.safetensors", | |
| "alpha": 8192.0, | |
| "base_model_name_or_path": "IFM/K2-Horizon-7B", | |
| "checkpoint_dir": "/mnt/weka/shrd/k2m/lingjie.chen/diffusion_train_runs/k2v3_phase2sft2500_stage5phase2_multispan64k_b8_tvonly_loraalpha8192_gbs128_8node_seed42_2274744/checkpoints/checkpoint_0000597", | |
| "converted_at_unix": 1788315252, | |
| "dropout": 0.05, | |
| "dtype_counts": { | |
| "torch.float32": 504 | |
| }, | |
| "lossless_exact_tensor_compare": true, | |
| "merged_tp_size": 1, | |
| "num_tensors": 504, | |
| "original_alpha": 8192.0, | |
| "original_rank": 128, | |
| "output_dir": "/mnt/weka/shrd/k2m/lingjie.chen/eval_models/k2v3p2_s5p2_ckpt0597_a8192_peft", | |
| "rank": 128, | |
| "rank_expansion": "exact_tp_block_diagonal", | |
| "qk_row_order": "hf_rotary_permuted_from_native_xllm", | |
| "query_heads": 32, | |
| "kv_heads": 8, | |
| "qk_roundtrip_validated": true, | |
| "save_dtype": "float32", | |
| "sample_shapes": { | |
| "model.layers.0.mlp.down_proj.lora_A.weight": [ | |
| 128, | |
| 12288 | |
| ], | |
| "model.layers.0.mlp.down_proj.lora_B.weight": [ | |
| 4096, | |
| 128 | |
| ], | |
| "model.layers.0.mlp.gate_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.mlp.gate_proj.lora_B.weight": [ | |
| 12288, | |
| 128 | |
| ], | |
| "model.layers.0.mlp.up_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.mlp.up_proj.lora_B.weight": [ | |
| 12288, | |
| 128 | |
| ], | |
| "model.layers.0.self_attn.k_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.self_attn.k_proj.lora_B.weight": [ | |
| 1024, | |
| 128 | |
| ], | |
| "model.layers.0.self_attn.o_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.self_attn.o_proj.lora_B.weight": [ | |
| 4096, | |
| 128 | |
| ], | |
| "model.layers.0.self_attn.q_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.self_attn.q_proj.lora_B.weight": [ | |
| 4096, | |
| 128 | |
| ], | |
| "model.layers.0.self_attn.v_proj.lora_A.weight": [ | |
| 128, | |
| 4096 | |
| ], | |
| "model.layers.0.self_attn.v_proj.lora_B.weight": [ | |
| 1024, | |
| 128 | |
| ], | |
| "model.layers.1.mlp.down_proj.lora_A.weight": [ | |
| 128, | |
| 12288 | |
| ], | |
| "model.layers.1.mlp.down_proj.lora_B.weight": [ | |
| 4096, | |
| 128 | |
| ] | |
| }, | |
| "source_config": "/mnt/weka/shrd/k2m/lingjie.chen/diffusion_train_runs/k2v3_phase2sft2500_stage5phase2_multispan64k_b8_tvonly_loraalpha8192_gbs128_8node_seed42_2274744/checkpoints/checkpoint_0000597/config.json", | |
| "target_modules": [ | |
| "q_proj", | |
| "k_proj", | |
| "v_proj", | |
| "o_proj", | |
| "gate_proj", | |
| "down_proj", | |
| "up_proj" | |
| ] | |
| } | |