Instructions to use trl-internal-testing/tiny-Glm5NextForConditionalGeneration with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use trl-internal-testing/tiny-Glm5NextForConditionalGeneration with Transformers:

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

pipe = pipeline("image-text-to-text", model="trl-internal-testing/tiny-Glm5NextForConditionalGeneration")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("trl-internal-testing/tiny-Glm5NextForConditionalGeneration")
model = AutoModelForMultimodalLM.from_pretrained("trl-internal-testing/tiny-Glm5NextForConditionalGeneration", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use trl-internal-testing/tiny-Glm5NextForConditionalGeneration with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "trl-internal-testing/tiny-Glm5NextForConditionalGeneration"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "trl-internal-testing/tiny-Glm5NextForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/trl-internal-testing/tiny-Glm5NextForConditionalGeneration

SGLang

How to use trl-internal-testing/tiny-Glm5NextForConditionalGeneration 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 "trl-internal-testing/tiny-Glm5NextForConditionalGeneration" \
    --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": "trl-internal-testing/tiny-Glm5NextForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "trl-internal-testing/tiny-Glm5NextForConditionalGeneration" \
        --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": "trl-internal-testing/tiny-Glm5NextForConditionalGeneration",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use trl-internal-testing/tiny-Glm5NextForConditionalGeneration with Docker Model Runner:
```
docker model run hf.co/trl-internal-testing/tiny-Glm5NextForConditionalGeneration
```

Upload Glm5NextForConditionalGeneration

by albertvillanova HF Staff - opened 1 day ago

base: refs/heads/main

←

from: refs/pr/3

Discussion Files changed

+180

-7

Files changed (3) hide show

config.json +177 -4
generation_config.json +1 -1
model.safetensors +2 -2

config.json CHANGED Viewed

@@ -7,11 +7,185 @@
   "image_start_token_id": 154830,
   "image_token_id": 154854,
   "model_type": "glm5_next",
   "text_config": {
     "attention_bias": false,
     "attention_dropout": 0.0,
     "bos_token_id": null,
-    "dtype": "bfloat16",
     "eos_token_id": [
       154820,
       154827,
@@ -45,7 +219,7 @@
       "linear_attention",
       "linear_attention",
       "linear_attention",
-      "deepseek_sparse_attention"
     ],
     "linear_attn_config": {
       "full_attn_layers": [
@@ -105,7 +279,7 @@
     "vocab_size": 154880
   },
   "tie_word_embeddings": false,
-  "transformers_version": "5.17.0",
   "video_end_token_id": 154833,
   "video_start_token_id": 154832,
   "video_token_id": 154855,
@@ -113,7 +287,6 @@
     "attention_bias": true,
     "attention_dropout": 0.0,
     "depth": 2,
-    "dtype": "bfloat16",
     "hidden_act": "silu",
     "hidden_size": 16,
     "image_size": 448,

   "image_start_token_id": 154830,
   "image_token_id": 154854,
   "model_type": "glm5_next",
+  "quantization_config": {
+    "activation_scheme": "dynamic",
+    "fmt": "e4m3",
+    "modules_to_not_convert": [
+      "attn_mha",
+      "attn_mqa",
+      "dt_bias",
+      "hyper_connection",
+      "lm_head",
+      "mapping_proj",
+      "model.embed_tokens",
+      "model.layers.0.hc_attn_base",
+      "model.layers.0.hc_attn_fn",
+      "model.layers.0.hc_attn_scale",
+      "model.layers.0.hc_ffn_base",
+      "model.layers.0.hc_ffn_fn",
+      "model.layers.0.hc_ffn_scale",
+      "model.layers.0.input_layernorm",
+      "model.layers.0.post_attention_layernorm",
+      "model.layers.0.self_attn.A_log",
+      "model.layers.0.self_attn.b_proj",
+      "model.layers.0.self_attn.dt_bias",
+      "model.layers.0.self_attn.f_a_proj",
+      "model.layers.0.self_attn.f_b_proj",
+      "model.layers.0.self_attn.fused_qkvbfg_a_proj",
+      "model.layers.0.self_attn.g_a_proj",
+      "model.layers.0.self_attn.g_b_proj",
+      "model.layers.0.self_attn.k_conv1d",
+      "model.layers.0.self_attn.k_proj",
+      "model.layers.0.self_attn.o_norm",
+      "model.layers.0.self_attn.o_proj",
+      "model.layers.0.self_attn.q_conv1d",
+      "model.layers.0.self_attn.q_proj",
+      "model.layers.0.self_attn.qkv_proj",
+      "model.layers.0.self_attn.v_conv1d",
+      "model.layers.0.self_attn.v_proj",
+      "model.layers.1.hc_attn_base",
+      "model.layers.1.hc_attn_fn",
+      "model.layers.1.hc_attn_scale",
+      "model.layers.1.hc_ffn_base",
+      "model.layers.1.hc_ffn_fn",
+      "model.layers.1.hc_ffn_scale",
+      "model.layers.1.input_layernorm",
+      "model.layers.1.post_attention_layernorm",
+      "model.layers.1.self_attn.A_log",
+      "model.layers.1.self_attn.b_proj",
+      "model.layers.1.self_attn.dt_bias",
+      "model.layers.1.self_attn.f_a_proj",
+      "model.layers.1.self_attn.f_b_proj",
+      "model.layers.1.self_attn.fused_qkvbfg_a_proj",
+      "model.layers.1.self_attn.g_a_proj",
+      "model.layers.1.self_attn.g_b_proj",
+      "model.layers.1.self_attn.k_conv1d",
+      "model.layers.1.self_attn.k_proj",
+      "model.layers.1.self_attn.o_norm",
+      "model.layers.1.self_attn.o_proj",
+      "model.layers.1.self_attn.q_conv1d",
+      "model.layers.1.self_attn.q_proj",
+      "model.layers.1.self_attn.qkv_proj",
+      "model.layers.1.self_attn.v_conv1d",
+      "model.layers.1.self_attn.v_proj",
+      "model.layers.2.hc_attn_base",
+      "model.layers.2.hc_attn_fn",
+      "model.layers.2.hc_attn_scale",
+      "model.layers.2.hc_ffn_base",
+      "model.layers.2.hc_ffn_fn",
+      "model.layers.2.hc_ffn_scale",
+      "model.layers.2.input_layernorm",
+      "model.layers.2.post_attention_layernorm",
+      "model.layers.2.self_attn.A_log",
+      "model.layers.2.self_attn.b_proj",
+      "model.layers.2.self_attn.dt_bias",
+      "model.layers.2.self_attn.f_a_proj",
+      "model.layers.2.self_attn.f_b_proj",
+      "model.layers.2.self_attn.fused_qkvbfg_a_proj",
+      "model.layers.2.self_attn.g_a_proj",
+      "model.layers.2.self_attn.g_b_proj",
+      "model.layers.2.self_attn.k_conv1d",
+      "model.layers.2.self_attn.k_proj",
+      "model.layers.2.self_attn.o_norm",
+      "model.layers.2.self_attn.o_proj",
+      "model.layers.2.self_attn.q_conv1d",
+      "model.layers.2.self_attn.q_proj",
+      "model.layers.2.self_attn.qkv_proj",
+      "model.layers.2.self_attn.v_conv1d",
+      "model.layers.2.self_attn.v_proj",
+      "model.layers.3.hc_attn_base",
+      "model.layers.3.hc_attn_fn",
+      "model.layers.3.hc_attn_scale",
+      "model.layers.3.hc_ffn_base",
+      "model.layers.3.hc_ffn_fn",
+      "model.layers.3.hc_ffn_scale",
+      "model.layers.3.input_layernorm",
+      "model.layers.3.mlp.gate",
+      "model.layers.3.mlp.gate.e_score_correction_bias",
+      "model.layers.3.post_attention_layernorm",
+      "model.layers.3.self_attn.indexer.index_kpool_compress_ape",
+      "model.layers.3.self_attn.indexer.index_kpool_compress_gate",
+      "model.layers.3.self_attn.indexer.k_norm",
+      "model.layers.3.self_attn.indexer.k_norm.bias",
+      "model.layers.3.self_attn.indexer.weights_proj",
+      "model.layers.3.self_attn.indexer.wk",
+      "model.layers.3.self_attn.indexer.wq_b",
+      "model.layers.3.self_attn.kv_a_layernorm",
+      "model.layers.3.self_attn.kv_b_proj",
+      "model.layers.3.self_attn.q_a_layernorm",
+      "model.layers.4.eh_proj",
+      "model.layers.4.enorm",
+      "model.layers.4.hnorm",
+      "model.layers.4.input_layernorm",
+      "model.layers.4.mlp.gate",
+      "model.layers.4.mlp.gate.e_score_correction_bias",
+      "model.layers.4.post_attention_layernorm",
+      "model.layers.4.self_attn.indexer.index_kpool_compress_ape",
+      "model.layers.4.self_attn.indexer.index_kpool_compress_gate",
+      "model.layers.4.self_attn.indexer.k_norm",
+      "model.layers.4.self_attn.indexer.k_norm.bias",
+      "model.layers.4.self_attn.indexer.weights_proj",
+      "model.layers.4.self_attn.indexer.wk",
+      "model.layers.4.self_attn.indexer.wq_b",
+      "model.layers.4.self_attn.kv_a_layernorm",
+      "model.layers.4.self_attn.kv_b_proj",
+      "model.layers.4.self_attn.q_a_layernorm",
+      "model.layers.4.shared_head.norm",
+      "model.norm",
+      "model.visual",
+      "router",
+      "visual",
+      "visual.blocks.0.attn.k_norm",
+      "visual.blocks.0.attn.proj",
+      "visual.blocks.0.attn.proj.bias",
+      "visual.blocks.0.attn.q_norm",
+      "visual.blocks.0.attn.qkv",
+      "visual.blocks.0.attn.qkv.bias",
+      "visual.blocks.0.mlp.down_proj",
+      "visual.blocks.0.mlp.down_proj.bias",
+      "visual.blocks.0.mlp.gate_proj",
+      "visual.blocks.0.mlp.gate_proj.bias",
+      "visual.blocks.0.mlp.up_proj",
+      "visual.blocks.0.mlp.up_proj.bias",
+      "visual.blocks.0.norm1",
+      "visual.blocks.0.norm2",
+      "visual.blocks.1.attn.k_norm",
+      "visual.blocks.1.attn.proj",
+      "visual.blocks.1.attn.proj.bias",
+      "visual.blocks.1.attn.q_norm",
+      "visual.blocks.1.attn.qkv",
+      "visual.blocks.1.attn.qkv.bias",
+      "visual.blocks.1.mlp.down_proj",
+      "visual.blocks.1.mlp.down_proj.bias",
+      "visual.blocks.1.mlp.gate_proj",
+      "visual.blocks.1.mlp.gate_proj.bias",
+      "visual.blocks.1.mlp.up_proj",
+      "visual.blocks.1.mlp.up_proj.bias",
+      "visual.blocks.1.norm1",
+      "visual.blocks.1.norm2",
+      "visual.downsample",
+      "visual.downsample.bias",
+      "visual.merger.down_proj",
+      "visual.merger.gate_proj",
+      "visual.merger.post_projection_norm",
+      "visual.merger.post_projection_norm.bias",
+      "visual.merger.proj",
+      "visual.merger.up_proj",
+      "visual.patch_embed.proj",
+      "visual.patch_embed.proj.bias",
+      "visual.post_layernorm",
+      "weights_proj"
+    ],
+    "quant_method": "fp8",
+    "weight_block_size": [
+      32,
+      32
+    ]
+  },
   "text_config": {
     "attention_bias": false,
     "attention_dropout": 0.0,
     "bos_token_id": null,
     "eos_token_id": [
       154820,
       154827,
       "linear_attention",
       "linear_attention",
       "linear_attention",
+      "indexed_attention"
     ],
     "linear_attn_config": {
       "full_attn_layers": [
     "vocab_size": 154880
   },
   "tie_word_embeddings": false,
+  "transformers_version": "5.23.0",
   "video_end_token_id": 154833,
   "video_start_token_id": 154832,
   "video_token_id": 154855,
     "attention_bias": true,
     "attention_dropout": 0.0,
     "depth": 2,
     "hidden_act": "silu",
     "hidden_size": 16,
     "image_size": 448,

generation_config.json CHANGED Viewed

@@ -7,5 +7,5 @@
   ],
   "pad_token_id": 154820,
   "temperature": 1.0,
-  "transformers_version": "5.17.0"
 }

   ],
   "pad_token_id": 154820,
   "temperature": 1.0,
+  "transformers_version": "5.23.0"
 }

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d2ce2a09050c4ebe531fc502524a2ade07549bb48e7d011e21d6e8b4397a14e0
-size 20082984

 version https://git-lfs.github.com/spec/v1
+oid sha256:87f1f1bdd919166e3737201dbaf3896f23882957f8782bb5a1c4184f344e0c86
+size 20042732