Instructions to use tiny-random/glm-5.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tiny-random/glm-5.2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="tiny-random/glm-5.2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tiny-random/glm-5.2")
model = AutoModelForCausalLM.from_pretrained("tiny-random/glm-5.2", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tiny-random/glm-5.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tiny-random/glm-5.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tiny-random/glm-5.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tiny-random/glm-5.2

SGLang

How to use tiny-random/glm-5.2 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 "tiny-random/glm-5.2" \
    --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": "tiny-random/glm-5.2",
		"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 "tiny-random/glm-5.2" \
        --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": "tiny-random/glm-5.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tiny-random/glm-5.2 with Docker Model Runner:
```
docker model run hf.co/tiny-random/glm-5.2
```

yujiepan commited on 30 days ago

Commit

cfdcfbb

verified ·

1 Parent(s): 5554360

Upload folder using huggingface_hub

Browse files

Files changed (3) hide show

README.md +10 -3
config.json +6 -1
model.safetensors +2 -2

README.md CHANGED Viewed

@@ -57,7 +57,7 @@ model = AutoModelForCausalLM.from_pretrained(
     dtype=torch.bfloat16,
     device_map=device,
 )
-generated_ids = model.generate(input_ids, max_new_tokens=8)
 output_text = tokenizer.decode(generated_ids[0][input_ids.shape[1]:])
 print(output_text)
 ```
@@ -85,7 +85,7 @@ from transformers import (
 )
 source_model_id = "zai-org/GLM-5.2"
-save_folder = "/tmp/tiny-random/glm-52"
 processor = AutoProcessor.from_pretrained(
     source_model_id, trust_remote_code=True)
@@ -100,6 +100,7 @@ config_json.update({
     "hidden_size": 8,
     "index_n_heads": 4,
     "indexer_types": ['full'] + ['shared'] * 3,
     "intermediate_size": 32,
     "moe_intermediate_size": 32,
     "num_hidden_layers": 4,
@@ -149,7 +150,7 @@ model.model.layers.append(nn.ModuleDict(dict(
     hnorm=nn.RMSNorm(config.hidden_size),
     input_layernorm=nn.RMSNorm(config.hidden_size),
     post_attention_layernorm=nn.RMSNorm(config.hidden_size),
-    self_attn=deepcopy(model.model.layers[1].self_attn),
     mlp=deepcopy(model.model.layers[1].mlp),
 )))
 for i in range(1, len(model.model.layers)):
@@ -237,6 +238,12 @@ GlmMoeDsaForCausalLM(
           (kv_a_layernorm): GlmMoeDsaRMSNorm((512,), eps=1e-06)
           (kv_b_proj): Linear(in_features=512, out_features=3584, bias=False)
           (o_proj): Linear(in_features=2048, out_features=8, bias=False)
         )
         (mlp): GlmMoeDsaMoE(
           (experts): GlmMoeDsaExperts(

     dtype=torch.bfloat16,
     device_map=device,
 )
+generated_ids = model.generate(input_ids, max_new_tokens=8)  # pyright: ignore[reportAttributeAccessIssue]
 output_text = tokenizer.decode(generated_ids[0][input_ids.shape[1]:])
 print(output_text)
 ```
 )
 source_model_id = "zai-org/GLM-5.2"
+save_folder = "/tmp/tiny-random/glm-52"  # pyright: ignore[reportUnusedExpression]
 processor = AutoProcessor.from_pretrained(
     source_model_id, trust_remote_code=True)
     "hidden_size": 8,
     "index_n_heads": 4,
     "indexer_types": ['full'] + ['shared'] * 3,
+    "index_topk_pattern": ['F'] + ['S'] * 3,
     "intermediate_size": 32,
     "moe_intermediate_size": 32,
     "num_hidden_layers": 4,
     hnorm=nn.RMSNorm(config.hidden_size),
     input_layernorm=nn.RMSNorm(config.hidden_size),
     post_attention_layernorm=nn.RMSNorm(config.hidden_size),
+    self_attn=deepcopy(model.model.layers[0].self_attn),
     mlp=deepcopy(model.model.layers[1].mlp),
 )))
 for i in range(1, len(model.model.layers)):
           (kv_a_layernorm): GlmMoeDsaRMSNorm((512,), eps=1e-06)
           (kv_b_proj): Linear(in_features=512, out_features=3584, bias=False)
           (o_proj): Linear(in_features=2048, out_features=8, bias=False)
+          (indexer): GlmMoeDsaIndexer(
+            (wq_b): Linear(in_features=32, out_features=512, bias=False)
+            (wk): Linear(in_features=8, out_features=128, bias=False)
+            (k_norm): LayerNorm((128,), eps=1e-06, elementwise_affine=True)
+            (weights_proj): Linear(in_features=8, out_features=4, bias=False)
+          )
         )
         (mlp): GlmMoeDsaMoE(
           (experts): GlmMoeDsaExperts(

config.json CHANGED Viewed

@@ -22,7 +22,12 @@
   "index_skip_topk_offset": 3,
   "index_topk": 2048,
   "index_topk_freq": 4,
-  "index_topk_pattern": null,
   "indexer_rope_interleave": true,
   "indexer_types": [
     "full",

   "index_skip_topk_offset": 3,
   "index_topk": 2048,
   "index_topk_freq": 4,
+  "index_topk_pattern": [
+    "F",
+    "S",
+    "S",
+    "S"
+  ],
   "indexer_rope_interleave": true,
   "indexer_types": [
     "full",

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:18db746dbdfb49ec5e6a211a7fa2223b7bc0011e33cd8916f43a5f1435cd9f24
-size 26180088

 version https://git-lfs.github.com/spec/v1
+oid sha256:151d405ef695562d64c615d24d4f0cbf0aec60d7b563c545ceb41484959a0744
+size 26216056