Triangle104 commited on
Commit
1c618fb
•
1 Parent(s): 0a8dc70

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +84 -0
README.md CHANGED
@@ -135,6 +135,90 @@ model-index:
135
  This model was converted to GGUF format from [`ibm-granite/granite-3.0-8b-instruct`](https://huggingface.co/ibm-granite/granite-3.0-8b-instruct) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space.
136
  Refer to the [original model card](https://huggingface.co/ibm-granite/granite-3.0-8b-instruct) for more details on the model.
137
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
138
  ## Use with llama.cpp
139
  Install llama.cpp through brew (works on Mac and Linux)
140
 
 
135
  This model was converted to GGUF format from [`ibm-granite/granite-3.0-8b-instruct`](https://huggingface.co/ibm-granite/granite-3.0-8b-instruct) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space.
136
  Refer to the [original model card](https://huggingface.co/ibm-granite/granite-3.0-8b-instruct) for more details on the model.
137
 
138
+ ---
139
+ Model Summary: Granite-3.0-8B-Instruct is a 8B parameter model finetuned from Granite-3.0-8B-Base using a combination of open source instruction datasets with permissive license and internally collected synthetic datasets. This model is developed using a diverse set of techniques with a structured chat format, including supervised finetuning, model alignment using reinforcement learning, and model merging.
140
+
141
+ Developers: Granite Team, IBM
142
+ GitHub Repository: ibm-granite/granite-3.0-language-models
143
+ Website: Granite Docs
144
+ Paper: Granite 3.0 Language Models
145
+ Release Date: October 21st, 2024
146
+ License: Apache 2.0
147
+
148
+ Supported Languages: English, German, Spanish, French, Japanese, Portuguese, Arabic, Czech, Italian, Korean, Dutch, and Chinese. Users may finetune Granite 3.0 models for languages beyond these 12 languages.
149
+
150
+ Intended use: The model is designed to respond to general instructions and can be used to build AI assistants for multiple domains, including business applications.
151
+
152
+ Capabilities
153
+
154
+ Summarization
155
+ Text classification
156
+ Text extraction
157
+ Question-answering
158
+ Retrieval Augmented Generation (RAG)
159
+ Code related tasks
160
+ Function-calling tasks
161
+ Multilingual dialog use cases
162
+
163
+ Generation: This is a simple example of how to use Granite-3.0-8B-Instruct model.
164
+
165
+ Install the following libraries:
166
+
167
+ pip install torch torchvision torchaudio
168
+ pip install accelerate
169
+ pip install transformers
170
+
171
+ Then, copy the snippet from the section that is relevant for your use case.
172
+
173
+ import torch
174
+ from transformers import AutoModelForCausalLM, AutoTokenizer
175
+
176
+ device = "auto"
177
+ model_path = "ibm-granite/granite-3.0-8b-instruct"
178
+ tokenizer = AutoTokenizer.from_pretrained(model_path)
179
+ # drop device_map if running on CPU
180
+ model = AutoModelForCausalLM.from_pretrained(model_path, device_map=device)
181
+ model.eval()
182
+ # change input text as desired
183
+ chat = [
184
+ { "role": "user", "content": "Please list one IBM Research laboratory located in the United States. You should only output its name and location." },
185
+ ]
186
+ chat = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
187
+ # tokenize the text
188
+ input_tokens = tokenizer(chat, return_tensors="pt").to(device)
189
+ # generate output tokens
190
+ output = model.generate(**input_tokens,
191
+ max_new_tokens=100)
192
+ # decode output tokens into text
193
+ output = tokenizer.batch_decode(output)
194
+ # print output
195
+ print(output)
196
+
197
+ Model Architecture: Granite-3.0-8B-Instruct is based on a decoder-only dense transformer architecture. Core components of this architecture are: GQA and RoPE, MLP with SwiGLU, RMSNorm, and shared input/output embeddings.
198
+ Model 2B Dense 8B Dense 1B MoE 3B MoE
199
+ Embedding size 2048 4096 1024 1536
200
+ Number of layers 40 40 24 32
201
+ Attention head size 64 128 64 64
202
+ Number of attention heads 32 32 16 24
203
+ Number of KV heads 8 8 8 8
204
+ MLP hidden size 8192 12800 512 512
205
+ MLP activation SwiGLU SwiGLU SwiGLU SwiGLU
206
+ Number of Experts — — 32 40
207
+ MoE TopK — — 8 8
208
+ Initialization std 0.1 0.1 0.1 0.1
209
+ Sequence Length 4096 4096 4096 4096
210
+ Position Embedding RoPE RoPE RoPE RoPE
211
+ # Paremeters 2.5B 8.1B 1.3B 3.3B
212
+ # Active Parameters 2.5B 8.1B 400M 800M
213
+ # Training tokens 12T 12T 10T 10T
214
+
215
+ Training Data: Overall, our SFT data is largely comprised of three key sources: (1) publicly available datasets with permissive license, (2) internal synthetic data targeting specific capabilities, and (3) very small amounts of human-curated data. A detailed attribution of datasets can be found in the Granite Technical Report and Accompanying Author List.
216
+
217
+ Infrastructure: We train Granite 3.0 Language Models using IBM's super computing cluster, Blue Vela, which is outfitted with NVIDIA H100 GPUs. This cluster provides a scalable and efficient infrastructure for training our models over thousands of GPUs while minimizing environmental impact by utilizing 100% renewable energy sources.
218
+
219
+ Ethical Considerations and Limitations: Granite 3.0 Instruct Models are primarily finetuned using instruction-response pairs mostly in English, but also multilingual data covering eleven languages. Although this model can handle multilingual dialog use cases, its performance might not be similar to English tasks. In such case, introducing a small number of examples (few-shot) can help the model in generating more accurate outputs. While this model has been aligned by keeping safety in consideration, the model may in some cases produce inaccurate, biased, or unsafe responses to user prompts. So we urge the community to use this model with proper safety testing and tuning tailored for their specific tasks.
220
+
221
+ ---
222
  ## Use with llama.cpp
223
  Install llama.cpp through brew (works on Mac and Linux)
224