sarvam-105b-FP8-dynamic

Model Overview

• Model Architecture: sarvamai/sarvam-105b
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws).
• Version: 1.0
• Model Developers: RedHatAI

This model is a quantized version of sarvamai/sarvam-105b. It was evaluated on several tasks to assess its quality in comparison to the unquantized model.

Model Optimizations

This model was obtained by quantizing the weights and activations of sarvamai/sarvam-105b to FP8 data type, ready for inference with vLLM.

Only the weights and activations of the linear operators within transformers blocks are quantized using LLM Compressor.

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend.

1. Install vLLM from main:

uv pip install -U git+https://github.com/vllm-project/vllm.git \
  --extra-index-url https://wheels.vllm.ai/nightly \
  --no-deps \
  --no-cache

2. Run using vLLM

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "RedHatAI/sarvam-105b-FP8-dynamic"
number_gpus = 1

sampling_params = SamplingParams(temperature=0.6, top_p=0.9, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompts = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)

llm = LLM(model=model_id, tensor_parallel_size=number_gpus)

outputs = llm.generate(prompts, sampling_params)

generated_text = outputs[0].outputs[0].text
print(generated_text)

vLLM also supports OpenAI-compatible serving. See the documentation for more details.

Creation

This model was created by applying LLM Compressor, as presented in the code snippet below.

Creation details

Install specific llm-compression version:

uv pip install git+https://github.com/vllm-project/llm-compressor.git
uv pip install --upgrade torchvision --break-system-packages --no-cache

from compressed_tensors.offload import dispatch_model
from transformers import AutoModelForCausalLM, AutoTokenizer

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier

MODEL_ID = "sarvamai/sarvam-105b"

# Load model.
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype="auto", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp8 with per channel via ptq
#   * quantize the activations to fp8 with dynamic per token
recipe = QuantizationModifier(
    targets="Linear", scheme="FP8_DYNAMIC", ignore=["lm_head"]
)

# Apply quantization.
oneshot(model=model, recipe=recipe)

# Confirm generations of the quantized model look sane.
print("========== SAMPLE GENERATION ==============")
dispatch_model(model)
input_ids = tokenizer("Hello my name is", return_tensors="pt").input_ids.to(
    model.device
)
output = model.generate(input_ids, max_new_tokens=20)
print(tokenizer.decode(output[0]))
print("==========================================")

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-Dynamic"
model.save_pretrained(SAVE_DIR)
tokenizer.save_pretrained(SAVE_DIR)

Evaluation

This model was evaluated on the well-known text benchmarks using lm-evaluation-harness.

  lm_eval \
    --model vllm \
    --model_args pretrained="RedHatAI/sarvam-105b-FP8-Dynamic",dtype=auto,add_bos_token=True,max_model_len=16384,tensor_parallel_size=2,gpu_memory_utilization=0.8,enable_chunked_prefill=True,trust_remote_code=True \
    --tasks openllm \
    --write_out \
    --batch_size auto \
    --show_config

Accuracy

Benchmark	sarvamai/sarvam-105b	RedHatAI/sarvam-105b-FP8-Dynamic	Recovery (%)
BBH (exact_match)	80.86	79.93	98.84%
GSM8K (strict-match)	84.38	85.37	101.17%
GSM8K (flexible-extract)	84.61	85.90	101.52%
IFEval (inst_level_strict_acc)	50.84	51.08	100.47%
MMLU-Pro (exact_match)	57.40	57.25	99.74%
ARC-Challenge (acc)	65.70	66.72	101.56%
HellaSwag (acc)	63.57	63.52	99.92%
MMLU (acc)	77.59	77.56	99.96%
TruthfulQA MC2 (acc)	51.21	51.64	100.85%
Winogrande (acc)	76.32	76.40	100.10%