Create README.md

README.md

@@ -0,0 +1,81 @@
+---
+license: apache-2.0
+language:
+- en
+tags:
+- retrieval
+- instructions
+- reranking
+---
+
+# Model Summary
+
+> FollowIR-7B is an instruction-tuned language model to be used for reranking in retrieval. It is Mistral-7B-Instruct-v0.2 fine-tuned on instruction retrieval data.
+
+- **Repository:** [ContextualAI/gritlm](https://github.com/ContextualAI/gritlm)
+- **Paper:** https://arxiv.org/abs/2402.09906
+- **Logs:** https://wandb.ai/muennighoff/gritlm/runs/0uui712t/overview
+- **Script:** https://github.com/ContextualAI/gritlm/blob/main/scripts/training/train_gritlm_7b.sh
+
+| Model | Description |
+|-------|-------------|
+| [GritLM 7B](https://hf.co/GritLM/GritLM-7B) | Mistral 7B finetuned using GRIT |
+| [GritLM 8x7B](https://hf.co/GritLM/GritLM-8x7B) | Mixtral 8x7B finetuned using GRIT |
+
+# Use
+
+Below is an example to compute the similarity score of a query-document pair
+```python
+model_name = "jhu-clsp/FollowIR-7B"
+model = AutoModelForCausalLM.from_pretrained(
+    model_name
+).cuda()
+tokenizer = AutoTokenizer.from_pretrained(
+    model_name, padding_side="left"
+)
+tokenizer.pad_token = tokenizer.eos_token
+tokenizer.padding_side = "left"
+token_false_id = tokenizer.get_vocab()["false"]
+token_true_id = tokenizer.get_vocab()["true"]
+max_length = min(2048, tokenizer.model_max_length)
+
+template = """<s> [INST] You are an expert Google searcher, whose job is to determine if the following document is relevant to the query (true/false). Answer using only one word, one of those two choices.
+
+Query: {query}
+Document: {text}
+Relevant (only output one word, either "true" or "false"): [/INST] """
+
+
+prompts = [
+    template.format(query=query, text=text) for (query, text) in zip([query] * 2, passages)
+]
+tokens = tokenizer(
+    prompts,
+    padding=True,
+    truncation=True,
+    return_tensors="pt",
+    max_length=max_length,
+    pad_to_multiple_of=None,
+)
+
+if "token_type_ids" in tokens:
+    del tokens["token_type_ids"]
+
+# move to cuda if desired
+for key in tokens:
+    tokens[key] = tokens[key].cuda()
+
+# calculate the scores
+batch_scores = model(**tokens).logits[:, -1, :]
+true_vector = batch_scores[:, token_true_id]
+false_vector = batch_scores[:, token_false_id]
+batch_scores = torch.stack([false_vector, true_vector], dim=1)
+batch_scores = torch.nn.functional.log_softmax(batch_scores, dim=1)
+scores = batch_scores[:, 1].exp().tolist()
+```
+
+# Citation
+
+```bibtex
+TODO
+```