fix typo

README.md

@@ -125,7 +125,7 @@ Please cite the paper if you refer to our model or code or data or paper from Wi
 ```
 
 
-❗To commen concern about dataset:
+❗To common concern about dataset:
 
 Recently, there have been clear changes in the open-source policy and regulations of our overall organization's code, data, and models.
 Despite this, we have still worked hard to obtain opening the weights of the model first, but the data involves stricter auditing and is in review with our legal team .
@@ -222,7 +222,7 @@ The following table provides a comprehensive comparison of WizardLMs and several
 | WizardCoder-15B  1.0 | **57.3**     |
 
 ## Call for Feedbacks
-We welcome everyone to use your professional and difficult instructions to evaluate WizardLM, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the the next version of WizardLM. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
+We welcome everyone to use your professional and difficult instructions to evaluate WizardLM, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the next version of WizardLM. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
 
 
 

WizardCoder/README.md

@@ -85,7 +85,7 @@ The following table clearly demonstrates that our **WizardCoder** exhibits a sub
 ❗**Note: The above table conducts a comprehensive comparison of our **WizardCoder** with other models on the HumanEval and MBPP benchmarks. We adhere to the approach outlined in previous studies by generating **20 samples** for each problem to estimate the pass@1 score and evaluate with the same [code](https://github.com/openai/human-eval/tree/master). The scores of GPT4 and GPT3.5 reported by [OpenAI](https://openai.com/research/gpt-4) are 67.0 and 48.1 (maybe these are the early version GPT4&3.5).**
 
 ## Call for Feedbacks
-We welcome everyone to use your professional and difficult instructions to evaluate WizardCoder, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the the next version of WizardCoder. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
+We welcome everyone to use your professional and difficult instructions to evaluate WizardCoder, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the next version of WizardCoder. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
 
 ## Unofficial Video Introductions
 Thanks to the enthusiastic friends, their video introductions are more lively and interesting.
@@ -159,7 +159,7 @@ deepspeed train_wizardcoder.py \
 
 ## Inference
 
-We provide the decoding script for WizardCoder, which reads a input file and generates corresponding responses for each sample, and finally consolidates them into an output file.
+We provide the decoding script for WizardCoder, which reads an input file and generates corresponding responses for each sample, and finally consolidates them into an output file.
 
 You can specify `base_model`, `input_data_path` and `output_data_path` in `src\inference_wizardcoder.py` to set the decoding model, path of input file and path of output file.
 

WizardLM/README.md

@@ -89,7 +89,7 @@ The following table provides a comprehensive comparison of WizardLMs and several
 | WizardLM-30B  1.0   | **37.8**         |
 
 ## Call for Feedbacks
-We welcome everyone to use your professional and difficult instructions to evaluate WizardLM, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the the next version of WizardLM. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
+We welcome everyone to use your professional and difficult instructions to evaluate WizardLM, and show us examples of poor performance and your suggestions in the [issue discussion](https://github.com/nlpxucan/WizardLM/issues) area. We are focusing on improving the Evol-Instruct now and hope to relieve existing weaknesses and issues in the next version of WizardLM. After that, we will open the code and pipeline of up-to-date Evol-Instruct algorithm and work with you together to improve it.
 
 ## Unofficial Video Introductions
 Thanks to the enthusiastic friends, their video introductions are more lively and interesting.
@@ -202,7 +202,7 @@ See [Distributed Fine-tuning](./doc/distributed_finetune.md)
 
 **NOTE:** The **WizardLM-13B-1.0** and **Wizard-7B** use different prompt at the beginning of the conversation!
 
-We provide the decoding script for WizardLM, which reads a input file and generates corresponding responses for each sample, and finally consolidates them into an output file.
+We provide the decoding script for WizardLM, which reads an input file and generates corresponding responses for each sample, and finally consolidates them into an output file.
 
 You can specify `base_model`, `input_data_path` and `output_data_path` in src\inference_wizardlm.py or src\infer_wizardlm13b.py to set the decoding model, path of input file and path of output file.
 

WizardLM/doc/distributed_finetune.md

@@ -3,9 +3,9 @@ We've conducted distributed fine tune experiment on our WizardLM utilizing origi
 To reproduce our experiments, we provided the steps and system configuration here.
 
 ## Steps
-We assume you have worker-0, worker-1, worker-2 which are GPU nodes to be used for training and they could ssh into each other via private key. We assume worker-0 is the master node here, which has a opened port MASTER_PORT that worker-1 and worker-2 can directly access and it has a MASTER_IP that other nodes can access.
+We assume you have worker-0, worker-1, worker-2 which are GPU nodes to be used for training and they could ssh into each other via private key. We assume worker-0 is the master node here, which has an opened port MASTER_PORT that worker-1 and worker-2 can directly access and it has a MASTER_IP that other nodes can access.
 
-In each worker, configure your enviorment using the instructions in Llama-X. Different workers should use the same absolute path in your data, output, code folder and they should be exactly the same configuration.
+In each worker, configure your environment using the instructions in Llama-X. Different workers should use the same absolute path in your data, output, code folder and they should be exactly the same configuration.
 
 After that, we need to change the hostfile config(*/path/to/Llama-X/src/configs/hostfile*) in each node, and add each worker into it, assuming 8 GPUs on each worker:
 ```bash
@@ -14,7 +14,7 @@ worker-1 slots=8
 worker-2 slots=8
 ```
 
-And since there might be some NCCL commuication problem considering the complexity of every cluster, we recommend use this config:
+And since there might be some NCCL communication problem considering the complexity of every cluster, we recommend use this config:
 ```bash
 NCCL_DEBUG=INFO
 NCCL_ASYNC_ERROR_HANDLING=1
@@ -71,4 +71,4 @@ NCCL_ASYNC_ERROR_HANDLING=1
 NCCL_IB_DISABLE=1
 NCCL_SOCKET_IFNAME=ens9f1
 ```
-NCCL_SOCKET_IFNAME needs to be changed to your worker's actual newtwork interface name, using *ifconfig* to find out.
+NCCL_SOCKET_IFNAME needs to be changed to your worker's actual network interface name, using *ifconfig* to find out.

WizardMath/README.md

@@ -214,7 +214,7 @@ Recently, there have been clear changes in the open-source policy and regulation
 
 <h2 id="inference">Inference</h2>
 
-We provide the decoding script for WizardMath, which reads a input file and generates corresponding responses for each sample, and finally calculate the score.
+We provide the decoding script for WizardMath, which reads an input file and generates corresponding responses for each sample, and finally calculate the score.
 
 ###  Install inference environment :
 Note: We used vllm for inference which can speed up inference and save time. Please refer to the official github [vllm](https://github.com/vllm-project/vllm/tree/main) for questions about vllm installation.

WizardMath/inference/MATH_inference.py

@@ -59,7 +59,7 @@ def test_hendrycks_math(model, data_path, start=0, end=MAX_INT, batch_size=1, te
         "Write a response that appropriately completes the request.\n\n"
         "### Instruction:\n{instruction}\n\n### Response: Let's think step by step."
     )
-    print('promt =====', problem_prompt)
+    print('prompt =====', problem_prompt)
     with open(data_path, "r+", encoding="utf8") as f:
         for idx, item in enumerate(jsonlines.Reader(f)):
             temp_instr = problem_prompt.format(instruction=item["instruction"])
@@ -71,12 +71,12 @@ def test_hendrycks_math(model, data_path, start=0, end=MAX_INT, batch_size=1, te
     print('total length ===', len(hendrycks_math_ins))
     hendrycks_math_ins = hendrycks_math_ins[start:end]
     hendrycks_math_answers = hendrycks_math_answers[start:end]
-    print('lenght ====', len(hendrycks_math_ins))
+    print('length ====', len(hendrycks_math_ins))
     batch_hendrycks_math_ins = batch_data(hendrycks_math_ins, batch_size=batch_size)
 
     stop_tokens = ["Instruction:", "Instruction", "Response:", "Response"]
     sampling_params = SamplingParams(temperature=0, top_p=1, max_tokens=2048, stop=stop_tokens)
-    print('sampleing =====', sampling_params)
+    print('sampling =====', sampling_params)
     llm = LLM(model=model,tensor_parallel_size=tensor_parallel_size)
     res_completions = []
     for idx, (prompt, prompt_answer) in enumerate(zip(batch_hendrycks_math_ins, hendrycks_math_answers)):

WizardMath/inference/gsm8k_inference.py

@@ -73,7 +73,7 @@ def gsm8k_test(model, data_path, start=0, end=MAX_INT, batch_size=1, tensor_para
         "Write a response that appropriately completes the request.\n\n"
         "### Instruction:\n{instruction}\n\n### Response: Let's think step by step."
     )
-    print('promt =====', problem_prompt)
+    print('prompt =====', problem_prompt)
     with open(data_path,"r+", encoding="utf8") as f:
         for idx, item in enumerate(jsonlines.Reader(f)):
             temp_instr = problem_prompt.format(instruction=item["question"])
@@ -84,12 +84,12 @@ def gsm8k_test(model, data_path, start=0, end=MAX_INT, batch_size=1, tensor_para
 
     gsm8k_ins = gsm8k_ins[start:end]
     gsm8k_answers = gsm8k_answers[start:end]
-    print('lenght ====', len(gsm8k_ins))
+    print('length ====', len(gsm8k_ins))
     batch_gsm8k_ins = batch_data(gsm8k_ins, batch_size=batch_size)
 
     stop_tokens = ["Instruction:", "Instruction", "Response:", "Response"]
     sampling_params = SamplingParams(temperature=0, top_p=1, max_tokens=1024, stop=stop_tokens)
-    print('sampleing =====', sampling_params)
+    print('sampling =====', sampling_params)
     llm = LLM(model=model,tensor_parallel_size=tensor_parallel_size)
     result = []
     res_completions = []

WizardMath/inference/util.py

@@ -75,7 +75,7 @@ def _clean_numbers(string):
     num_prev_digits = 0
     new_string = ""
     for i, c in enumerate(string):
-        # isdigit() doesnt work here because of weird unicode chars.
+        # isdigit() doesn't work here because of weird unicode chars.
         if c in {'1', '2', '3', '4', '5', '6', '7', '8', '9', '0'}:
             num_prev_digits += 1
         else: