🔥 We’re thrilled to announce the release of KAT-Dev-72B-Exp, our latest and most powerful model yet!
🔥 You can now try our strongest proprietary coder model KAT-Coder directly on the StreamLake platform for free.
KAT-Dev-72B-Exp is an open-source 72B-parameter model for software engineering tasks.
On SWE-Bench Verified, KAT-Dev-72B-Exp achieves 74.6% accuracy ⚡ — when evaluated strictly with the SWE-agent scaffold.
KAT-Dev-72B-Exp is the experimental reinforcement-learning version of the KAT-Coder model. Through this open-source release, we aim to reveal the technical innovations behind KAT-Coder’s large-scale RL to developers and researchers.
We rewrote the attention kernel and redesigned the training engine for shared prefix trajectories to achieve highly efficient RL training, especially for scaffolds leveraging context management.
Furthermore, to prevent exploration collapse observed in RL training, we reshaped advantage distribution based on pass rates: amplifying the advantage scale of highly exploratory groups while reducing that of low-exploration ones.
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "KAT-Dev-72B-Exp"
# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto"
)
# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
# conduct text completion
generated_ids = model.generate(
**model_inputs,
max_new_tokens=65536
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
content = tokenizer.decode(output_ids, skip_special_tokens=True)
print("content:", content)
temperature: 0.6
max_turns: 150
history_processors.n: 100
For full settings please refer to inference.yaml
model-00006-of-00031.safetensors 5.0 GBmodel-00010-of-00031.safetensors 5.0 GBmodel-00014-of-00031.safetensors 5.0 GBmodel-00018-of-00031.safetensors 5.0 GBmodel-00022-of-00031.safetensors 5.0 GBmodel-00026-of-00031.safetensors 5.0 GBmodel-00002-of-00031.safetensors 5.0 GBmodel-00008-of-00031.safetensors 4.8 GBmodel-00009-of-00031.safetensors 4.8 GBmodel-00012-of-00031.safetensors 4.8 GBmodel-00013-of-00031.safetensors 4.8 GBmodel-00016-of-00031.safetensors 4.8 GBmodel-00017-of-00031.safetensors 4.8 GBmodel-00020-of-00031.safetensors 4.8 GBmodel-00021-of-00031.safetensors 4.8 GBmodel-00024-of-00031.safetensors 4.8 GBmodel-00025-of-00031.safetensors 4.8 GBmodel-00028-of-00031.safetensors 4.8 GBmodel-00029-of-00031.safetensors 4.8 GBmodel-00005-of-00031.safetensors 4.8 GBmodel-00007-of-00031.safetensors 4.8 GBmodel-00011-of-00031.safetensors 4.8 GBmodel-00015-of-00031.safetensors 4.8 GBmodel-00019-of-00031.safetensors 4.8 GBmodel-00023-of-00031.safetensors 4.8 GBmodel-00027-of-00031.safetensors 4.8 GBmodel-00004-of-00031.safetensors 4.8 GBmodel-00003-of-00031.safetensors 4.8 GBmodel-00001-of-00031.safetensors 4.5 GBmodel-00030-of-00031.safetensors 3.2 GBmodel-00031-of-00031.safetensors 2.5 GBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo Kwaipilot/KAT-Dev-72B-Exp ./model-folder