14.7 GB

Ctrl+K

1 contributor

update mtp.safetensors: MTP v6 epoch 16 (best), trained 2026-03-22, KL loss on train_40k_greedy_v2, val_acc=64.3%, online k=1=67.8%, k=3 speedup=1.36x@c=1

4fa1152 verified 7 days ago

.gitattributes

1.58 kB
Upload model.safetensors.index.json with huggingface_hub 21 days ago
.gitignore

38 Bytes
apilog_k25_mtp/output_v3/checkpoints/7 18 days ago
README.md

24 Bytes
initial commit 21 days ago
chat_template.jinja

3.76 kB
Upload chat_template.jinja with huggingface_hub 21 days ago
config.json

5.35 kB
Upload config.json with huggingface_hub 21 days ago
configuration_deepseek.py

10.6 kB
Upload configuration_deepseek.py with huggingface_hub 21 days ago
configuration_kimi_k25.py

5.44 kB
Upload configuration_kimi_k25.py with huggingface_hub 21 days ago
convert_mtp_fp4_to_int4.py

6.07 kB
Upload convert_mtp_fp4_to_int4.py with huggingface_hub 21 days ago
convert_mtp_fp8_to_int4.py

4.25 kB
Upload convert_mtp_fp8_to_int4.py with huggingface_hub 21 days ago
generation_config.json

52 Bytes
Upload generation_config.json with huggingface_hub 21 days ago
kimi_k25_processor.py

6.91 kB
Upload kimi_k25_processor.py with huggingface_hub 21 days ago
kimi_k25_vision_processing.py

10 kB
Upload kimi_k25_vision_processing.py with huggingface_hub 21 days ago
media_utils.py

13.2 kB
Upload media_utils.py with huggingface_hub 21 days ago
model.safetensors.index.json

23.9 MB
xet

Upload model.safetensors.index.json with huggingface_hub 21 days ago
modeling_deepseek.py

78.6 kB
apilog_k25_mtp/output_v3/checkpoints/7 18 days ago
modeling_kimi_k25.py

50.9 kB
Upload modeling_kimi_k25.py with huggingface_hub 21 days ago
mtp.safetensors

14.7 GB
xet

update mtp.safetensors: MTP v6 epoch 16 (best), trained 2026-03-22, KL loss on train_40k_greedy_v2, val_acc=64.3%, online k=1=67.8%, k=3 speedup=1.36x@c=1 7 days ago
patch_vllm.py

3.25 kB
Upload patch_vllm.py with huggingface_hub 21 days ago
preprocessor_config.json

742 Bytes
Upload preprocessor_config.json with huggingface_hub 21 days ago
tiktoken.model

2.8 MB
feat: add 4th patch (eagle.py image_token_index fallback), FP4→INT4 converter, fix tiktoken.model LFS pointer, update weight index 21 days ago
tokenization_kimi.py

13.4 kB
Upload tokenization_kimi.py with huggingface_hub 21 days ago
tokenizer_config.json

4.86 kB
Upload tokenizer_config.json with huggingface_hub 21 days ago
tool_declaration_ts.py

16.5 kB
Upload tool_declaration_ts.py with huggingface_hub 21 days ago