Text Generation
Transformers
Safetensors
English
Chinese
glm_moe_dsa
glm
mixture-of-experts
fp8
expert-pruning
man
conversational
Instructions to use destr8803/GLM-5.3-224E-FP8-MAN with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use destr8803/GLM-5.3-224E-FP8-MAN with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="destr8803/GLM-5.3-224E-FP8-MAN") 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("destr8803/GLM-5.3-224E-FP8-MAN") model = AutoModelForCausalLM.from_pretrained("destr8803/GLM-5.3-224E-FP8-MAN", 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 destr8803/GLM-5.3-224E-FP8-MAN with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "destr8803/GLM-5.3-224E-FP8-MAN" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "destr8803/GLM-5.3-224E-FP8-MAN", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/destr8803/GLM-5.3-224E-FP8-MAN
- SGLang
How to use destr8803/GLM-5.3-224E-FP8-MAN 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 "destr8803/GLM-5.3-224E-FP8-MAN" \ --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": "destr8803/GLM-5.3-224E-FP8-MAN", "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 "destr8803/GLM-5.3-224E-FP8-MAN" \ --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": "destr8803/GLM-5.3-224E-FP8-MAN", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use destr8803/GLM-5.3-224E-FP8-MAN with Docker Model Runner:
docker model run hf.co/destr8803/GLM-5.3-224E-FP8-MAN
Download prune-report.json from destr8803/GLM-5.3-224E-FP8-MAN: direct link, hf CLI and curl.
- Browser
- Download file 4.35 kB
-
https://huggingface.co/destr8803/GLM-5.3-224E-FP8-MAN/resolve/main/prune-report.json
- Command line
-
hf download hf://destr8803/GLM-5.3-224E-FP8-MAN/prune-report.json
-
curl -L -o prune-report.json https://huggingface.co/destr8803/GLM-5.3-224E-FP8-MAN/resolve/main/prune-report.json
4.35 kB
| { | |
| "base_model": "zai-org/GLM-5.3", | |
| "calibration": { | |
| "axes": { | |
| "agentic": { | |
| "samples": 648, | |
| "tokens": 497302 | |
| }, | |
| "code": { | |
| "samples": 912, | |
| "tokens": 500387 | |
| }, | |
| "general": { | |
| "samples": 1128, | |
| "tokens": 495244 | |
| }, | |
| "reasoning": { | |
| "samples": 160, | |
| "tokens": 490789 | |
| } | |
| }, | |
| "corpus_sha256": "27bdf438f05019d65655c2f0db7ef0b7f994d3c2b250abe66244f30af986a372", | |
| "packing": false, | |
| "samples": 2848, | |
| "seed": 42, | |
| "source_datasets": [ | |
| "allenai/c4[en]", | |
| "theblackcat102/evol-codealpaca-v1", | |
| "open-r1/Mixture-of-Thoughts[code,math,science]", | |
| "glaiveai/glaive-function-calling-v2", | |
| "SWE-bench/SWE-smith-trajectories[tool]" | |
| ], | |
| "tokens": 1983722, | |
| "truncation": false, | |
| "world_size": 8 | |
| }, | |
| "checkpoint": { | |
| "config_sha256": "546c69d5e6b79f848598f0cc5ad7689bf5033cd3995e34341918ae7224158c7e", | |
| "index_sha256": "a8193674cab604ca7b8af9bfc5ff1db10a058a832282017c8aa345a77c3c99c6", | |
| "payload_bytes": 663759907008, | |
| "precision": "FP8 weights; BF16 KV cache runtime-qualified", | |
| "shards": 141, | |
| "tensors": 104037 | |
| }, | |
| "evaluation": { | |
| "behavior": { | |
| "passed": 12, | |
| "reasoning_efforts": [ | |
| "low", | |
| "high", | |
| "max" | |
| ], | |
| "total": 12 | |
| }, | |
| "comparison_report_sha256": "e2acf585eed9108fe502d8573eca46a700c534134721ae997bc46baec2e8e297", | |
| "functional": { | |
| "passed": 4, | |
| "total": 4 | |
| }, | |
| "gpqa_diamond": { | |
| "accuracy_percent": 97.5, | |
| "bootstrap_95_clustered_by_question": [ | |
| 92.5, | |
| 100.0 | |
| ], | |
| "completed_requests": 40, | |
| "correct": 39, | |
| "dataset_sha256": "41d1213cd7a4998605a26c2798500652572007161b3a92817ba46b35befcd305", | |
| "dataset_url": "https://openaipublic.blob.core.windows.net/simple-evals/gpqa_diamond.csv", | |
| "failed_requests": 0, | |
| "malformed_answers": 0, | |
| "max_tokens": 120000, | |
| "prompt_sha256": "d4d85337bbd54659d12c8da7edc8b854d151b7327cf7574dc1062d753b03c8a6", | |
| "question_count": 40, | |
| "question_set_sha256": "c5c24ae147b5f771e3b5ca5552119fa1340e479073c23adad46b22f21ebc3476", | |
| "reasoning_effort": "max (server default)", | |
| "run_fingerprint": "a8a8737cfeae2bd96a87df27b1620ad523dcac5a1dc341b108a1d524f943ca4e", | |
| "seed": 0, | |
| "temperature": 1.0, | |
| "top_p": 0.95 | |
| }, | |
| "runtime": { | |
| "accelerators": "8x Intel Gaudi2", | |
| "aggregate_16x1024_tokens_per_second": 237.03, | |
| "expert_parallel_size": 8, | |
| "kv_cache_dtype": "BF16", | |
| "max_model_len": 131072, | |
| "single_stream_tokens_per_second": 19.716830071821388, | |
| "tensor_parallel_size": 8 | |
| } | |
| }, | |
| "format_version": 3, | |
| "license": "glm-5.3; see LICENSE", | |
| "limitations": [ | |
| "The checkpoint is a 12.5% expert-pruned derivative without retraining.", | |
| "Calibration was English-heavy and did not include an explicit Spanish axis.", | |
| "GPQA validation covered a deterministic 40-question subset, not the full benchmark.", | |
| "Gaudi2 TP8/EP8 was qualified; other accelerator backends were not qualified." | |
| ], | |
| "pruning": { | |
| "activation_outlier_protection": { | |
| "candidate_count": 7, | |
| "candidate_layers": [ | |
| 77 | |
| ], | |
| "enabled": true, | |
| "threshold": "max_abs > max(global_p99.5, global_max/10)" | |
| }, | |
| "criterion": "S(1,0,1)=mean(l2(unweighted_expert_output)) over routed tokens", | |
| "criterion_id": "s101_man", | |
| "keep_map_sha256": "f5abe51855b4d7975773bf640128c3077fcf8f7aaa3c54f1c892b55304e2444b", | |
| "method": "MAN", | |
| "mtp_policy": "reuse_layer_77_keep_ids_for_base_first_delivery", | |
| "reduction_percent": 12.5, | |
| "retraining": false, | |
| "source_experts_per_sparse_layer": 256, | |
| "target_experts_per_sparse_layer": 224, | |
| "top_k": 8 | |
| }, | |
| "repository": "destr8803/GLM-5.3-224E-FP8-MAN", | |
| "source": { | |
| "config_sha256": "3ac72612095574542f7fff847ada8e59d9199dd8af44bdf625d7e02615572e69", | |
| "index_sha256": "e0fe7f28c1f853d4824e4d796374e3dacf1fe470988773952c79b063768134bf", | |
| "routed_experts_per_sparse_layer": 256 | |
| }, | |
| "status": "runtime-qualified", | |
| "validation": { | |
| "checked_tensors": 104037, | |
| "errors": 0, | |
| "report_sha256": "0d547367304885ae4d7c66b44e7c446e0dcd7685ff0a085992604e865d23a391", | |
| "shards": 141, | |
| "status": "pass" | |
| } | |
| } | |