Lin-K76 commited on
Commit
2545394
1 Parent(s): 90ff7a0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +4 -4
README.md CHANGED
@@ -34,7 +34,7 @@ This optimization reduces the number of bits per parameter from 16 to 8, reducin
34
  Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-tensor quantization is applied, in which a single linear scaling maps the FP8 representations of the quantized weights and activations.
35
  [AutoFP8](https://github.com/neuralmagic/AutoFP8) is used for quantization with 512 sequences of UltraChat.
36
 
37
- <!-- ## Deployment
38
 
39
  ### Use with vLLM
40
 
@@ -46,7 +46,7 @@ from transformers import AutoTokenizer
46
 
47
  model_id = "neuralmagic/Mistral-Nemo-Instruct-2407-FP8"
48
 
49
- sampling_params = SamplingParams(temperature=0.6, top_p=0.9, max_tokens=256)
50
 
51
  tokenizer = AutoTokenizer.from_pretrained(model_id)
52
 
@@ -57,7 +57,7 @@ messages = [
57
 
58
  prompts = tokenizer.apply_chat_template(messages, tokenize=False)
59
 
60
- llm = LLM(model=model_id)
61
 
62
  outputs = llm.generate(prompts, sampling_params)
63
 
@@ -65,7 +65,7 @@ generated_text = outputs[0].outputs[0].text
65
  print(generated_text)
66
  ```
67
 
68
- vLLM aslo supports OpenAI-compatible serving. See the [documentation](https://docs.vllm.ai/en/latest/) for more details. -->
69
 
70
  ## Creation
71
 
 
34
  Only the weights and activations of the linear operators within transformers blocks are quantized. Symmetric per-tensor quantization is applied, in which a single linear scaling maps the FP8 representations of the quantized weights and activations.
35
  [AutoFP8](https://github.com/neuralmagic/AutoFP8) is used for quantization with 512 sequences of UltraChat.
36
 
37
+ ## Deployment
38
 
39
  ### Use with vLLM
40
 
 
46
 
47
  model_id = "neuralmagic/Mistral-Nemo-Instruct-2407-FP8"
48
 
49
+ sampling_params = SamplingParams(temperature=0.3, top_p=0.9, max_tokens=256)
50
 
51
  tokenizer = AutoTokenizer.from_pretrained(model_id)
52
 
 
57
 
58
  prompts = tokenizer.apply_chat_template(messages, tokenize=False)
59
 
60
+ llm = LLM(model=model_id, max_model_len=4096)
61
 
62
  outputs = llm.generate(prompts, sampling_params)
63
 
 
65
  print(generated_text)
66
  ```
67
 
68
+ vLLM aslo supports OpenAI-compatible serving. See the [documentation](https://docs.vllm.ai/en/latest/) for more details.
69
 
70
  ## Creation
71