Update README.md

README.md

@@ -1,3 +1,223 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+pipeline_tag: text-generation
+tags:
+- fp8
+- quantized
+- llm-compressor
+- compressed-tensors
+- red hat
+base_model:
+- ibm-granite/granite-4.0-h-tiny
+---
+
+
+# Granite-4.0-h-tiny
+
+## Model Overview
+- **Model Architecture:** GraniteMoeHybridForCausalLM
+  - **Input:** Text
+  - **Output:** Text
+- **Model Optimizations:**
+  - **Weight quantization:** FP8
+  - **Activation quantization:** FP8
+- **Release Date:** 
+- **Version:** 1.0
+- **Model Developers:**: Red Hat
+
+Quantized version of [ibm-granite/granite-4.0-h-tiny](https://huggingface.co/ibm-granite/granite-4.0-h-tiny).
+
+### Model Optimizations
+
+This model was obtained by quantizing the weights and activations of [ibm-granite/granite-4.0-h-tiny](https://huggingface.co/ibm-granite/granite-4.0-h-tiny) to FP8 data type.
+This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%.
+Only the weights and activations of the linear operators within transformers blocks of the language model are quantized. 
+
+## Deployment
+
+### Use with vLLM
+
+1. Install specific version:
+```
+uv pip install -U git+https://github.com/vllm-project/vllm.git \
+  --extra-index-url https://wheels.vllm.ai/nightly \
+  --no-deps \
+  --no-cache
+
+uv pip install compressed-tensors==0.12.3a20251114 --no-cache
+uv pip install --upgrade torchvision --break-system-packages --no-cache
+uv pip install cloudpickle msgspec zmq blake3 cachetools prometheus_client fastapi openai openai_harmony pybase64 llguidance diskcache xgrammar lm-format-enforcer partial-json-parser cbor2 einops gguf numba --no-cache
+
+```
+
+2. Initialize vLLM server:
+```
+vllm serve RedHatAI/granite-4.0-h-tiny-FP8-block --tensor_parallel_size 1
+```
+
+3. Send requests to the server:
+
+```python
+from openai import OpenAI
+
+# Modify OpenAI's API key and API base to use vLLM's API server.
+openai_api_key = "EMPTY"
+openai_api_base = "http://<your-server-host>:8000/v1"
+
+client = OpenAI(
+    api_key=openai_api_key,
+    base_url=openai_api_base,
+)
+
+model = "RedHatAI/granite-4.0-h-tiny-FP8-block"
+
+messages = [
+    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
+]
+
+
+outputs = client.chat.completions.create(
+    model=model,
+    messages=messages,
+)
+
+generated_text = outputs.choices[0].message.content
+print(generated_text)
+```
+
+## Creation
+
+This model was quantized using the [llm-compressor](https://github.com/vllm-project/llm-compressor) library as shown below.
+
+
+<details>
+  <summary>Creation details</summary>
+
+Install specific llm-compression version:
+```
+uv pip install git+https://github.com/vllm-project/llm-compressor.git@refs/pull/2001/head --no-cache
+uv pip install --upgrade torchvision --break-system-packages --no-cache
+```
+
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+from llmcompressor import oneshot
+from llmcompressor.modifiers.quantization import QuantizationModifier
+from llmcompressor.utils import dispatch_for_generation
+from llmcompressor.modeling import replace_modules_for_calibration
+from llmcompressor.modeling.granite4 import pack_3d_experts
+
+MODEL_ID = "ibm-granite/granite-4.0-h-tiny"
+
+model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
+
+model = replace_modules_for_calibration(model)
+
+ignore_lay = ["lm_head", "re:.*block_sparse_moe.router", "re:.*mamba.in_proj", "re:.*shared_mlp.input_linear"]
+
+recipe = QuantizationModifier(
+    targets=["Linear"],
+    scheme="FP8_BLOCK",
+    ignore=ignore_lay,
+)
+
+oneshot(model=model, recipe=recipe)
+
+print("========== SAMPLE GENERATION ==============")
+dispatch_for_generation(model)
+input_ids = tokenizer(
+    "Describe Large Language Model", return_tensors="pt"
+).input_ids.to(model.device)
+output = model.generate(input_ids, max_new_tokens=35)
+print(tokenizer.decode(output[0]))
+print("==========================================")
+
+SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-block"
+print(f"Saving to {SAVE_DIR}")
+
+model.save_pretrained(SAVE_DIR)
+tokenizer.save_pretrained(SAVE_DIR)
+pack_3d_experts(SAVE_DIR)
+```
+</details>
+
+
+## Evaluation
+
+
+The model was evaluated on the OpenLLM leaderboard task, using [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness).
+[vLLM](https://docs.vllm.ai/en/stable/) was used for all evaluations.
+
+<details>
+  <summary>Evaluation details</summary>
+
+  Install specific version:
+```
+uv pip install -U git+https://github.com/vllm-project/vllm.git \
+  --extra-index-url https://wheels.vllm.ai/nightly \
+  --no-deps \
+  --no-cache
+
+
+uv pip install compressed-tensors==0.12.3a20251114 --no-cache
+uv pip install --upgrade torchvision --break-system-packages --no-cache
+uv pip install cloudpickle msgspec zmq blake3 cachetools prometheus_client fastapi openai openai_harmony pybase64 llguidance diskcache xgrammar lm-format-enforcer partial-json-parser cbor2 einops gguf numba --no-cache
+```
+  
+  **Openllm V1**
+  ```
+  lm_eval \
+    --model vllm \
+    --model_args pretrained="RedHatAI/granite-4.0-h-tiny-FP8-block",dtype=auto,add_bos_token=True,max_model_len=16384,tensor_parallel_size=1,gpu_memory_utilization=0.9,enable_chunked_prefill=True,trust_remote_code=True \
+    --tasks openllm \
+    --write_out \
+    --batch_size auto \
+    --show_config
+  ```
+
+
+  **Openllm V2**  
+  ```
+  lm_eval \
+    --model vllm \
+    --model_args pretrained="RedHatAI/granite-4.0-h-tiny-FP8-block",dtype=auto,add_bos_token=False,max_model_len=16384,tensor_parallel_size=1,gpu_memory_utilization=0.7,disable_log_stats=True,enable_chunked_prefill=True,trust_remote_code=True \
+    --tasks leaderboard \
+    --apply_chat_template \
+    --fewshot_as_multiturn \
+    --write_out \
+    --batch_size auto \
+    --show_config
+  ```
+
+
+  **Coding Benchmarks**
+
+  ```
+  evalplus.evaluate --model "RedHatAI/granite-4.0-h-tiny-FP8-block" \
+                    --dataset "humaneval" \
+                    --backend vllm \
+                    --tp 1 \
+                    --greedy
+
+  evalplus.evaluate --model "RedHatAI/granite-4.0-h-tiny-FP8-block" \
+                  --dataset "mbpp" \
+                  --backend vllm \
+                  --tp 1 \
+                  --greedy
+
+  ```
+
+</details>
+
+
+<!-- <b>*</b> I/p Length = 2048, O/p Length = 2048, #Requests = 1024 -->
+
+
+
+### Accuracy Comparison
+
+
+